Fachbereich Mathematik und Naturwissenschaften Fachgruppe...

Fachbereich Mathematik und Naturwissenschaften

Fachgruppe Physik

Bergische Universitat Wuppertal

Oktober 2009

Topics in the Measurement of

Top Quark Events with ATLAS

Pixel Detector Optoelectronics,Track Impact Parameter Calibration,Acceptance Correction Methods

Dissertation zur Erlangung des Doktorgrades

vorgelegt von

Stephan A. Sandvoss

Diese Dissertation kann wie folgt zitiert werden: urn:nbn:de:hbz:468-20100160 [http://nbn-resolving.de/urn/resolver.pl?urn=urn%3Anbn%3Ade%3Ahbz%3A468-20100160]

To Objectivity, Reliability, Validity,and Teamwork

Uberblick

Diese Arbeit beschreibt Verfahren, die insbesondere zur Messung von Top-QuarkEreignissen mit dem ATLAS Detektor am Large Hadron Collider (LHC) desCERN angewendet werden konnen. Es wurden Beitrage zu drei wesentlichen Be-reichen erbracht: dem Detektoraufbau und seiner Uberprufung, der Datenkali-bration und der ersten physikalischen Analyse.

Der Pixel Detektor verfugt uber etwa 80 Millionen Auslesekanale und repra-sentiert damit mehr als 90% aller Auslesekanale des ATLAS Detektors. Es wurdeeine Methode zur Anwendung eines optischen Reflektometers entwickelt, mitder die Unversehrtheit der Lichtwellenleiterkabel, die die auslesende Optoelek-tronik außerhalb des Pixel Detektors mit derjenigen innerhalb verbinden, schnellgetestet werden kann. Damit wurde die Einsatzbereitschaft des Pixel Detektorsnach Produktion, Test und Installation dieser Ausleseelektronik (Back of CrateKarten) sichergestellt.

Prazisionsmessungen erfordern eine sorgfaltige Detektorkalibration. Hierzukann ein Verfahren beitragen, das die Spurstoßparameter von Monte Carlo (MC)Datensatzen an die Verteilungen von Daten anpaßt. Zunachst werden die Spur-und Jetrekonstruktion sowie das b-Tagging von Jets in ATLAS vorgestellt. DieStoßparameter von Spuren sind vor allem durch den innersten Detektor bestimmt,den Pixel Detektor. Die Genauigkeit der Spurstoßparameter hat großen Einflußauf die Resultate des b-Taggings und damit auf die Selektion und Analyse vonTop Quark Ereignissen.

Zu Anfang der Messungen wird es Abweichungen in den Stoßparameterver-teilungen zwischen simulierten und realen Daten geben, z.B. weil die relativeAnordnung von Detektorteilen nicht genau genug bekannt ist. Auf die Spurenmit negativen Stoßparametern gestutzt, die hauptsachlich von der intrinsischenDetektorauflosung abhangen, werden die Stoßparameter der simulierten Spurenangepaßt. Eine Implementierung im Athena Framework von ATLAS wurde auf-grund fehlender Kollisionsdaten mit zwei MC Top-Quark-Paar Datensatzen ver-schiedener Detektorgeometrien getestet. Mit der verfugbaren Statistik lassen sichVerbesserungen bei der Ubereinstimmung von simulierten mit (pseudo-) realenDaten fur Spuren mit Stoßparametern kleiner als 0.4 mm bis zu einem Faktor von4 erreichen.

Die Analyse der ersten Top-Quark Ereignisse dient sowohl der Kalibrationund Leistungsverbesserung des ATLAS Detektors als auch der Validierung bis-heriger Messungen und der Uberprufung theoretischer Vorhersagen bei Energien,die nie zuvor erreicht wurden. Ein weiteres Verfahren gewichtet die Vorhersagenvon Monte Carlo Generatoren fur Top-Quark Observablen, so daß sie mit Datenubereinstimmen. Bei Top-Quark Messungen lassen sich so systematische Unsi-cherheiten verringern, die durch unterschiedliche Modellierung zugrundeliegenderphysikalischer Prozesse verursacht werden. Zunachst werden alle Schritte der MCSimulation in ATLAS vorgestellt, von der Generierung eines Ereignisses uber dieSimulation im Detektor bis zu seiner Rekonstruktion.Die Vorhersagen von funf Monte Carlo Generatoren werden hinsichtlich der tota-len Selektionseffizienz von semileptonischen Top-Quark-Paar Ereignissen vergli-chen. Diese Effizienzen weichen wie einige Verteilungen der Selektionsobservablenund des Transversalimpulses des hadronisch zerfallenden Top Quarks um bis zu20% voneinander ab. Das fuhrt zu einer systematischen Unsicherheit der Messungaufgrund der Modellierung des zugrundeliegenden physikalischen Prozesses.Um diese Unsicherheit fur die ersten Daten zu verringern, rewichtet das entwik-kelte Verfahren die MC Ereignisse anhand einer einzigen gemessenen Verteilung:anhand des Transversalimpulses des hadronisch zerfallenden Top-Quarks. EineImplementierung und Anwendung im Athena Framework zeigt Verbesserungenvon bis zu einem Faktor 10, die systematische Unsicherheit in der totalen Selek-tionseffizienz fallt um mehr als den Faktor 4 von etwa 20% auf unter 5%.

Contents

Preface 1

1 Introduction 3

2 Theoretical Overview 52.1 The Standard Model of Particle Physics . . . . . . . . . . . . . . 52.2 Top Quark Physics . . . . . . . . . . . . . . . . . . . . . . . . . . 102.3 Monte Carlo Generators . . . . . . . . . . . . . . . . . . . . . . . 12

3 LHC and ATLAS 173.1 The Large Hadron Collider (LHC) . . . . . . . . . . . . . . . . . 173.2 The ATLAS Experiment . . . . . . . . . . . . . . . . . . . . . . . 20

3.2.1 The Muon System . . . . . . . . . . . . . . . . . . . . . . 253.2.2 The Calorimeters . . . . . . . . . . . . . . . . . . . . . . . 263.2.3 The Inner Detector . . . . . . . . . . . . . . . . . . . . . . 28

4 Optoelectronics of the Pixel Detector System 314.1 The Pixel Detector . . . . . . . . . . . . . . . . . . . . . . . . . . 314.2 The Pixel Detector Readout: The Optical Link . . . . . . . . . . 344.3 Production, Installation and Commissioning of Optoelectronics . . 35

4.3.1 Back Of Crate Cards . . . . . . . . . . . . . . . . . . . . . 354.3.2 Optical Cables . . . . . . . . . . . . . . . . . . . . . . . . 42

5 Event Reconstruction 515.1 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515.2 Vertexing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555.3 Electrons and Photons . . . . . . . . . . . . . . . . . . . . . . . . 565.4 Missing Transverse Energy . . . . . . . . . . . . . . . . . . . . . . 565.5 Jets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6 Tuning of Track Impact Parameters for b-Tagging 616.1 b-Tagging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616.2 Adjustment Strategy . . . . . . . . . . . . . . . . . . . . . . . . . 666.3 Performance and Tests . . . . . . . . . . . . . . . . . . . . . . . . 68

ii CONTENTS

6.4 Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

7 Systematic Uncertainties of Acceptance Corrections 757.1 Monte Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . 777.2 Total and Differential Efficiencies . . . . . . . . . . . . . . . . . . 807.3 Reweighting of MC Events from (Pseudo) Data . . . . . . . . . . 887.4 Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

8 Summary and Outlook 97

A Implementation/Plots of the Track Impact Parameter Tuning 99A.1 Implementation of the Track Impact Parameter Mapping Approach 99A.2 Plots of the Track Impact Parameter Mapping Approach . . . . . 101

B Implementation/Plots of the Reweighting of top pair Events 123B.1 Implementation of the Analysis and of the Reweighting . . . . . . 123B.2 Plots of the Analysis and of the Reweighting . . . . . . . . . . . . 124

List of Figures 139

List of Tables 141

Glossary 143

Bibliography 147

Preface

I am strongly inspired by Richard Feynman, his “Lectures on Physics” and manyother of his books. Therefore, I start off with these excerpts from his lectures,that are most important to me:“We must, incidentally, make it clear from the beginning that if a thing is not ascience, it is not necessarily bad. For example, love is not a science. So, if some-thing is said not to be a science, it does not mean that there is something wrongwith it; it just means that it is not a science.” [1, I, 3-1] “Poets say science takesaway from the beauty of the stars — mere globs of gas atoms. Nothing is ‘mere’.I too can see the stars on a desert night, and feel them...It does not do harm tothe mystery to know a little about it.” [1, I, 3-4] “In order to understand physicallaws, you must understand that they are all some kind of approximation. In thesame way, to define the mass of a single object is impossible, because there arenot any single, left-alone objects in the world—every object is a mixture of a lotof things, so we can deal with it only as a series of approximations and idealizia-tions. One may prefer a mathematical definition; but mathematical definitionscan never work in the real world.” [1, I, 12-1] “In its efforts to learn as much aspossible about nature, modern physics has found that certain things can never be‘known’ with certainty. Much of our knowledge must always remain uncertain.The most we can know is in terms of probabilities. This is the best descriptionof nature that one can give.” [1, I, 6-5] “The test of all knowledge is experiment.Experiment is the sole judge of scientific ‘truth’.” [1, I, 1-1]“Why repeat all this? Because there are new generations born every day. Becausethere are great ideas developed in the history of man, and these ideas do not lastunless they are passed purposely and clearly from generation to generation.” [2]

For any questions the author can be reached by sending an email [email protected].

Acknowledgements

I would like to thank my supervisor, Prof. Dr. Peter Mattig, to give me the op-portunity to do my PhD under his survey.

Many thanks to all members of the ATLAS Group of Wuppertal for theirfriendship and support.

Thanks to all colleagues of ATLAS and CERN who helped me according tocircumstances.

Stephan Sandvoss

Chapter 1

Introduction

Currently, several thousands of scientists from all over the world are preparing forthe first beam collisions at the Large Hadron Collider (LHC). This accelerator atthe European Center for Nuclear Research (CERN) in Geneva/Switzerland willprovide the opportunity to make precision measurements of top quark propertiesand eventually to discover the last fundamental particle proposed by the Stan-dard Model (SM) of particle physics: the Higgs particle. In order to do so, thedetectors, for example ATLAS, have to work reliably and be well understood, andsystematic uncertainties in the event reconstruction or in the physics modellingmust be reduced as much as possible.

In this thesis contributions to three fields were made: hardware production,installation and commissioning for the ATLAS pixel detector (Chapter 4), datacalibration (Chapter 6), and physics analysis preparation for first data (Chapter7). Each of these three fields is important for or strongly related to top quarkphysics.This thesis is organized as follows: A theoretical overview of particle physics,especially of the top quark sector, is given in Chapter 2.

Top quarks will be produced in the four interaction points of the LHC. Cen-tered around these points, where the two proton beams circling in opposite direc-tions collide, large detectors are installed: ALICE, ATLAS, CMS, LHCb. Theywill measure the products of the collisions, and hopefully some unexpected physicswill be discoverd. The main features of LHC and ATLAS are presented in Chap-ter 3.

A part of the work for this thesis was devoted to the production, installationand commissioning of optoelectronics for the innermost subdetector of ATLAS:the pixel detector. This detector is crucial for a precise reconstruction of thecollision events, especially for the accurate measurements of electrons, verticesand for b-(jet-)tagging. The pixel detector has more than 8 ·107 readout channels,these represent more than 90% of all ATLAS readout channels. The steeringand the data readout of the pixel detector is performed through an optical datatransmission line, the optical link. Chapter 4 overviews the pixel detector and the

4 Chapter 1. Introduction

optical link. In particular, the work for and the results of production, installationand commissioning of the corresponding optoelectronics are presented.

After detection and recording of the collision events the raw data is analyzed.Physics analyses like top quark mass measurements need high level reconstructedobjects such as jets or electrons. Therefore, a complete event reconstruction isperformed and the corresponding information is stored in separate files. Thereconstruction of all common high level objects is summarized in Chapter 5.

After their reconstruction, collision events are analyzed to find, for example,certain Higgs particle decays or to perform top quark precision measurements.In order to attain such goals, it is crucial to determine, whether jets originatefrom the fragmentation/hadronization of b-quarks. This identification method iscalled b-tagging, and the physics analyses mentioned above heavily rely on itsperformance. Track impact parameters are the leading ingredients for this pro-cedure, which are mainly determined by the pixel detector.Initially, there may be disagreements between track impact parameter distribu-tions of Monte Carlo (MC) and real data samples. Such systematic uncertaintiescan be caused for example by misaligned detector modules. In Chapter 6, amethod is presented, with which the MC track impact parameters can be tunedfrom data. Hence, agreement between simulation and data, and, therefore, asmaller systematic uncertainty shall be reached.

In Chapter 7, the predictions of MC generators for top quark pair produc-tion and decay are analyzed. Five common MC generators were used to simulatethe semileptonic decay channel of top-antitop events. The distributions of ob-servables, which are relevant to the selection and analysis of such events, arepresented and compared with each other.Furthermore, a method is presented that allows to reweight the Monte Carlo pre-dictions such that they agree better with data. In addition, the method providesan estimate of the systematic uncertainty caused by imperfections of the physicalmodel.

Finally, the results of this work are summarized, and an outlook to the forth-coming tasks is given in Chapter 8.

Chapter 2

Theoretical Overview

Top quark pair events of MC datasets are used in Chapter 6 and 7 to test themethods developed there and make some estimations for first data. Therefore,this chapter overviews the theory related to the top quark and MC generators.The top quark represents a fundamental particle of the Standard Model (SM) ofparticle physics. This model describes the behavior of matter at dimensions ofabout 10−15 − 10−18 m and is currently most supported by experiments.

2.1 The Standard Model of Particle Physics

At dimensions like or below the size of a molecule particles behave quantum-mechanically. In high energy physics, subatomic particles are used for scientificresearch, which usually travel at relativistic velocities near the speed of light. Asa consequence, theoretical calculations for and interpretations of measurementsof quantum particles have to be according to special relativity (Lorentz transfor-mations). The theory that is supported by most experiments of particle physicsand combines quantum effects and special relativity is a relativistic quantum fieldtheory. It assumes the conservation of quantities like energy, momentum, electriccharge and angular momentum for quantum particles.

Particles are characterized by their properties, among these are mass, elec-trical charge and intrinsic angular momentum, which is called spin. Hence, theyare important observables in high energy physics, for example in precision mea-surements of top quark characteristics. Some particle properties, or quantumnumbers, are additive, for example the momentum, other are multiplicative, forexample parity, which is linked to space reflection. There is for each particlean antiparticle, which has the opposite additive quantum numbers and the samemultiplicative quantum numbers. For example, the antitop quark t is the an-tiparticle of the top quark t.Particles are divided into two groups by regarding their intrinsic property spin:the ones that obey Fermi-Dirac statistics are called fermions, to which the top

6 Chapter 2. Theoretical Overview

quark belongs. Fermions have half-integral spin (in units of ~), whereas bosons,which represent the second group, have integral spin and obey Bose-Einsteinstatistics.

Currently, four fundamental forces are known to act between particles: thestrong, the electromagnetic, the weak and the gravitational force. The actionbetween particles can be described by the exchange of virtual particles, whichcannot be measured. These are called gauge bosons, and most of them can in-teract directly with each other.As illustrated in figure 2.1, the gravitation is at dimensions, which are typical

Figure 2.1: The relative strenghts of the four fundamental forces in dependence of the dis-tance. The current experimental limit is noted as well. [3]

of particle physics (< 10−15 m), by many orders of magnitudes weaker than thefirst three interactions. Therefore, gravitation is omitted in the Standard Modelof particle physics and is irrelevant at the LHC, except if extra dimensions exist.Up to now, there is neither any direct experimental evidence of a gauge boson of

2.1 The Standard Model of Particle Physics 7

gravitation (graviton) nor a satisfactory quantum mechanical theory of gravity.Hence, the classical theory of general relativity is still the most accurate descrip-tion of this force. Table 2.1 overviews the three interactions of the StandardModel and their gauge bosons, and table 2.2 summarizes the known fermions.

interaction/force strong electromagnetic weak

couplingcolor electrical weak

charge charge chargeapproximate

10−15 ∞ 10−17

range [m]gauge bosons 8 gluons (g) photon (γ) W+, W−, Z0

their mass0 0 80 , 80 , 91

[GeV/c2]

Table 2.1: The three fundamental forces of the Standard Model and their basic properties.

The corresponding force carrier particles and their masses (in units of GeV/c2 according toEinstein’s equivalence of mass and energy: E = m · c2) are noted as well.

Since all three forces of the Standard Model act on the top quark, they areshortly presented. Quantum Chromo-Dynamics (QCD), the theory of the strongforce, describes the interactions between so called color charges. There are threecolors, or quantum states, which are usually called “blue”, “green” and “red”, butthey have nothing to do with hue or frequency of light. Free observable particlesare color singlets (“white”), in other words, they have zero net color. Directlyobservable particles that are acted upon by the strong force are called hadrons.They have a substructure and consist of bound colored particles called partons,which are not directly observable.There are two types of partons: quarks (q) and gluons (g). Gluons are the gaugebosons of the strong force, and they have spin 1, no electric charge and no restmass. Gluons can interact directly with each other via the strong force, becausethey carry color charge. A color-neutral gluon was never observed, hence thereare only 8 different gluons, although 3 colors times 3 anticolors would result in 9independent combinations or states.Quarks are massive fermions with spin 1

2and carry a color charge of unit one.

An antiquark q carries an anticolor of unit one (“antiblue”, “antigreen” or “an-tired”) and has the opposite electric charge as the quark. Six types or flavorsof quarks are known: d(own)- and u(p)-, s(trange)- and c(harm)-, b(ottom)- andt(op)-quark. Their masses can be never measured directly in a well defined way,because free quarks have never been observed. 3 of the quark flavors — d, s andb — have an electric charge of −1

3each. Whereas, a u-, c- and t-quark has an

electric charge of +23

each. Except quarks all observable particles have integralcharges, that are multiples of the elementary charge, as we will see below.


Leptons 1. Generation 2. Generation 3. Generationflavor νe e νµ µ ντ τelectric

0 −1 0 −1 0 −1charge [e]approximate

< 3 · 10−6 0.511 < 0.19 105.7 < 18 1777mass [MeV/c2]forces el.mag. el.mag. el.mag.acted upon weak

Quarks 1. Generation 2. Generation 3. Generationflavor d u s c b telectric −1/3 +2/3 −1/3 +2/3 −1/3 +2/3charge [e]approximate

5 3 120 1250 4200 175000mass [MeV/c2]forces

strong, electromagnetic, weakacted upon

Table 2.2: The basic properties and the interactions of the elementary fermions (spin 1

2)

known from the Standard Model. Leptons and quarks are sorted into three generations. Thegravitation acts upon all particles.

Apart from their mass and life time, the d-, s- and b-quark have identical prop-erties, the same is true for the u-, c− and t-quark. Therefore, the quarks aregrouped into three pairs, ordered by mass: d- and u-quark are called first genera-tion quarks. s- and c− quark represent the second generation, b- and t-quark thethird. However, only the d- and u-quarks — together with gluons — may formstable hadrons.Up to now, two ways were observed, in which quarks are combined to form ahadron: mesons and baryons, both together represent the hadrons, which are thestrongly interacting particles. A combination of a quark q and an antiquark q iscalled meson, which is not stable and decays after some time. Its quark and anti-quark have corresponding color and anticolor, so that the meson is color-neutral.Three quarks, which are bound together by gluons, constitute a baryon. Eachof these quarks has a different color, so that the baryon is color-neutral. Thebest known baryons are the proton, which consists of 2 u- and 1 d-quark, andthe neutron, which consists of 1 u- and 2 d-quarks. Protons and neutrons arethe constituents (nucleons) of atomic nuclei. Usually, protons and neutrons arebound together in the nucleus by the strong force. However, not all combina-tions of protons and neutrons represent stable nuclei. Unstable nuclei decay orradiate particles, hence they are called radioactive. For example, nuclei decay, ifthe electrical repulsion between the protons, which carry an electrical charge ofunit 1, is stronger than the strong attraction between the nucleons.

2.1 The Standard Model of Particle Physics 9

The electromagnetic force explains the interactions between electric chargesand their interactions with light. According to Millikan’s experiment, electriccharges are positive (+) or negative (−) and an integral multiple of the elementarycharge e. A proton carries an electric charge of +e, an electron −e. Two chargesattract each other, if their signs are different, otherwise they repel. The theory ofclassical electrodynamics using static and dynamic fields was generalized to theQuantum Electro-Dynamics (QED). This theory describes electromagnetic effectsby probabilities for the emission, flow and absorption of photons (γ). These arethe gauge bosons of the electromagnetic force, which have spin 1 and no restmass. They cannot interact directly with each other via the electromagneticforce, because they have no electric charge. Furthermore, the photon is its ownantiparticle. The agreement between QED and experiments is very precise. Forexample, the value of the anomalous magnetic moment of the electron agreesbetween theory and experiment better than a part in a billion (109).

Finally, the weak force, which is the reason for the decay of the top quark,has 3 gauge bosons: W+, W−, Z0. They have spin 1 and rest masses of about80, 80 and 91 GeV/c2. The weak force acts upon hadrons and leptons, which arefermions not affected by the strong force, because they have no substructure ofcolored particles.There are three types or flavors of electrically charged leptons, which differ onlyin their rest mass and life time: e±1 (electron/positron), µ±1 (muon), τ±1 (tau).Only the lightest, the electron, is stable. Moreover, there are three flavors ofelectrically neutral leptons, which are called neutrinos: νe, νµ, ντ . Accordingto recent measurements, there are neutrino oscillations: a neutrino can make atransition from one flavor to any another, e.g. νe → νµ. This implies that at least2 neutrinos have a nonzero rest mass. Like quarks, leptons can be divided into 3generations, which have the same properties apart from rest mass and life time:e and νe represent the first, µ and νµ the second, τ and ντ the third generation.Since hadrons and leptons are affected by the weak force, there are 9 types of weaktransition processes of particles. The initial state can be hadronic, leptonic ormixed (semileptonic), that is, there are only hadrons, leptons or both. The sameis true for the final state. However, the total lepton and total baryon numberare conserved under the weak interaction as well as under the other interactions.Moreover, the weak force is responsible for flavor changes of leptons or quarks.All possible transitions from one quark flavor to another quark flavor can bedescribed by a corresponding combination of 9 probabilities, which represent theCabibbo-Kobayashi-Maskawa (CKM) matrix.The weak force explains also several effects as the beta decay or the proton-proton fusion in stars. Moreover, the weak force is the only interaction underwhich parity and CP-symmetry (symmetry of charge conjugation and parity) areviolated.

As it can be seen in figure 2.1, the electromagnetic and weak force have aboutthe same relative strength at dimensions below 10−18 m. This dimension corre-


sponds to an energy of Λ ≈ 246 GeV, the so called electroweak scale. Therefore,the theories of the electromagnetic and weak force are combined to the elec-troweak theory, which suggests one further boson: the (Standard Model) Higgsparticle. It is most likely that this particle completes the fundamental particlesof the Standard Model. The Higgs particle is predicted to be massive, electricallyneutral and to have spin 0. Though experimentalists search the Higgs boson al-ready for several decades, it has not been found yet. A mass below 114.4 GeV/c2

and between 160 and 170 GeV/c2 is excluded at 95% confidence level by directexperimental searches [4, 5]. The most recent particle accelerator, the LargeHadron Collider (LHC), will provide by proton-proton collisions searches in massranges of up to 14 TeV/c2. The main Higgs production mechanism at the LHCis the gluon fusion process for all possible Higgs masses. The cross section ofthis process is determined by the Higgs-fermion coupling. Since the top quarkmass is much larger than that of all other known fermions, the Higgs productionis dominated by the Higgs-top coupling. Therefore, a detailed understanding ofthe top quark is important for the Higgs sector. The LHC and one of its fourexperiments, ATLAS, are presented in Chapter 3.

It seems that the Higgs boson is the only missing part for a full validation ofthe Standard Model. However, there are several other puzzles and inconsistencies,which arise from different experiments and measurements. For example, due tothe observed neutrino oscillations mentioned above either a neutrinoless doublebeta decay or right-handed neutrinos should be observable. Moreover, neitherobservations of glueballs, which are hadrons containing no valence quark, nor ofpentaquarks, which are baryons containing four quarks and one antiquark, havebeen confirmed, though QCD allows such objects. Furthermore, neither the widemass range of the fermions (the so called hierarchy problem) nor the net excessof matter compared to antimatter by about 10 orders of magnitude (probablycaused by CP violation) are satisfactorily explained.Measurements in the top quark sector may play a major role in solving thementioned inconsistencies. Since the top quark is by far the heaviest knownfundamental fermion and thus quite near to the electroweak scale Λ ≈ 246 GeV,its properties may indicate new physics. The following section introduces the topquark sector.

2.2 Top Quark Physics

The experiments of the Tevatron [6] accelerator, CDF [7] and DØ [8], discoveredthe top quark. Many top quark properties were measured there, for instanceits mass (172.4 ± 1.2 GeV/c2 [9]). At the LHC, precision measurements withtop quarks will be performed with high statistics: The cross section of a tt-production in proton-proton collisions with design center of mass energy (14 TeV)will be about σ(tt) ≈ 800 pb [10], which corresponds to 10 tt-pairs per second

2.2 Top Quark Physics 11

at design luminosity (1034 1cm2·s). Such a production happens mainly by a gluon-

gluon fusion, which is illustrated by the three leading Feynman diagrams in figure2.2.

�g t

g g

t �ggt

t

t �gg t

t

t

Figure 2.2: Feynman diagrams of a top/antitop quark pair production by gluon-gluon fu-sion. [11]

The life time of the top quark (≈ 10−24 s) is shorter than the typical time, inwhich bound states of hadrons are formed by the strong interaction (≈ 10−23 s).Therefore, this process called hadronization is very unprobable for top quarks,and hadrons containing top quarks will most likely never be observed. Since allother quarks are much lighter than the top, its decay is the only observable decayof a quark in an unbound state.The decay of the top quark is reconstructed from detectable decay products. Inmore than 99% of the cases the top quark decays into a W boson and a bottomquark, so the other two possible transitions (to a strange or down quark) arenegligible. Since a W can decay into a charged lepton and its correspondingneutrino (≈ 33%) or into a quark-antiquark pair (≈ 67%) [12] the final state ofa tt-decay can be leptonic, hadronic or mixed (semileptonic), as illustrated infigure 2.3. Since free quarks do not exist, they form color-neutral, often unstablehadrons, which later decay into several particles. Usually, these are also unstable,

Figure 2.3: The final states of the decay of a top/antitop quark pair. [13]


what results in a long decay chain, until stable particles are left. Due to the strongproduction process at the parton level, most particles originating from a certainquark are concentrated in a small solid angle and measured as a single objectcalled jet. Hence, a hadronic final state is also called a jet state; vice versathe semileptonic state is called lepton+jets state. Since τ -leptons can decayhadronically, the branching ratios observed in a detector may be different fromthe theoretical values provided by table 2.3, if the τ is not reconstructed.

final state leptonic lepton+jets jetschannel ee µµ ττ eµ eτ µτ e+jets µ+jets τ+jets jetsBR [%] 1.2 1.2 1.2 2.4 2.4 2.4 14.7 14.7 14.7 45.1

Table 2.3: Branching ratios of the decay of a top-antitop quark pair in percent. [12]

In order to measure precisely top quark properties, an accurate event recon-struction is required. Ideally, the energies and momenta of all particles and jetsin the event are well measured and can be used for that purpose. However,neutrinos cannot be detected, because they interact neither strongly nor elec-tromagnetically. As a consequence, the momentum of the collision event in theplane transverse to the colliding particles seems to be nonzero. That is whyone or several neutrinos are also referred as missing transverse energy (Emiss

T or/ET). In order to understand such effects, to estimate the influence of backgroundprocesses and to compare experimental data with theoretical predictions, it isnecessary to simulate particle collision events. This is described in the next sec-tion.

2.3 Monte Carlo Generators

The experimental measurements of top quark properties at Tevatron were in-cluded into the theory, in order to make further predictions. But since no ana-lytical expressions are available for some quantities of experimental interest andsince the kinematics of single collision events shall be studied, they have to bepredicted by Monte Carlo generators. These are programs which use theoreticalor emperical models and to which data from experiments is given. The generatorssimulate collisions of high energetic particles, in which many new particles areproduced like illustrated in figure 2.4. As stated in [14], all Monte Carlo eventgenerators adhere to the paradigm to divide a collision event into stages followingthe diagram of figure 2.4 inside out:

• The hard (high energetic) scattering process of two partons, when twohadrons collide. The probability to resolve a parton i with the longitu-dinal momentum fraction xi inside a hadron by a momentum transfer of

2.3 Monte Carlo Generators 13

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Figure 2.4: Sketch of a proton-proton collision event simulated by Monte Carlo. The incomingprotons are drawn in dark blue. The dark red blob represents the hard scattering process oftwo colliding hadrons. QCD-Bremsstrahlung is drawn for the initial state (blue) and final state(red). The modelling of multiple parton interactions (underlying event) is shown in purple.The hadronization of partons is represented by light green ellipses. The decays of hadrons aredrawn in dark green. Finally, QED-Bremsstrahlung is sketched in yellow. [15]

Q2i between the incoming particle and the resolved parton is encoded in the

parton distribution function f(xi, Q2i ) (PDF). The scattering process of two

partons can be calculated by integrating the matrix element (probabilitydistribution) of the perturbation theory over a large phase space (momen-tum, spin). Usually, the calculation can be done within reasonable timeonly for matrix elements that represent the leading order (LO) or next-to-leading order (NLO) of perturbation theory. The leading order does notcontain any loop and is also called tree level. The next-to-leading orderinvolves one-loop and one-leg amplitudes.

• The QCD-Bremsstrahlung (gluons), which is emitted by accelerated coloredpartons. This leads to a cascade of emissions, which can be calculated bya parton shower approach. It uses perturbation theory (usually to leadinglogarithmic accuracy LLA) and stops the cascade by a cut off at a lowenergy scale (≈ 1 GeV & ΛQCD), where perturbation theory is not reliable.

• The hadronization, which confines colored partons into uncolored (“white”)


hadrons. This is not predictive from first principles, so phenomenologicalmodels are used, for example the string [16] or the cluster [17] model.

• The QED-Bremsstrahlung (photons), which is emitted by charged acceler-ated particles and mathematically identical with QCD-Bremsstrahlung.

• The underlying event, which is not well known, even often not clearly de-fined. It can include initial/final state showers, the hadronization of theremnants of the two collided particles, interactions between both remnants(multiple parton interactions), etc.

• The pile-up, which means that there is more than one interaction in eachbeam crossing. The number of soft or hard interactions per bunch crossingdepends on the luminosity.

There exist many different approaches for every stage, and each MC generator is acertain combination of these approaches. Table 2.4 gives an overview of the mainproperties of five common MC generators (Pythia [18], AcerMC [19], Sherpa [20],Herwig [17], MC@NLO [21]). More information about MC generators can befound in [14]. Due to the various used approaches, MC generators may provide

ModelMatrix Parton

HadronizationUnderlying

Element Shower Event

Pythia LO pT orderedstring

multiple

modelinteractions,

beam remnants

Herwig LOangular cluster

minimum-bias

ordered modellike UA5 [22] orvia JIMMY [23]

Sherpa LO pT orderedcluster or multiplevia Pythia interactions

AcerMC LO pT orderedvia Pythia via Pythiaor Herwig or Herwig

MC@NLO NLO pT ordered via Herwig via Herwig

Table 2.4: Basic properties of five common Monte Carlo generators.

different predictions for the same physical process. But experiments can decide,which predictions are more accurate for a certain physical process and a givenMC generator.

If the MC predictions differ from data and systematic errors are excluded,there are two alternatives to provide an agreement between theory (MC gener-ator) and experiment: First, the MC generators can be tuned, until simulation

2.3 Monte Carlo Generators 15

and measurement agree within statistical and systematic uncertainty. This canbe an elaborate and time consuming procedure. The alternative is to reweightthe MC distributions, so that they agree with data afterwards. Such an approachfor an analysis of the first data is presented in this thesis using the example ofobservables of top quark pair events. But first, we have a look on how particlesare accelerated, lead to collision and how such a collision is detected.

Chapter 3

LHC and ATLAS

3.1 The Large Hadron Collider (LHC)

The Large Hadron Collider (LHC) [24] started beam commissioning in Septem-ber 2008 at the laboratory of the European Organization for Nuclear Research(CERN) in Geneva. The LHC is located about 100 m underground in a largering tunnel of about 27 Km circumference. It is partly in France and partly inSwitzerland, between the Jura mountains and the airport of Geneva. The LHCcan accelerate protons or ions.

At four points of the ring the two LHC vacuum beam pipes intersect. Here,the charged particles of the two beams travelling in opposite directions collide.In such collisions lots of new particles are produced, which fly from the collisionpoint into all directions. By measuring these particles with a large detector,that is arranged around the interaction point, it can be concluded how matterbehaves at the highest energies or the smallest dimensions. Figure 3.1 shows aschematic view of the LHC and its four experiments ALICE (A Lhc Ion ColliderExperiment) [25], ATLAS (A Toroidal Lhc ApparatuS) [26], CMS (CompactMuon Solenoid) [27] and LHCb (LHC beauty experiment) [28].

A collider system of several linear accelerators and two ring accelerators (Pro-ton Synchrotron (PS) [29] and Super Proton Synchrotron (SPS) [30]) providesthe LHC with the beam particles. If these are protons, they are delivered withan energy of 450 GeV. The LHC can accelerate them up to an energy of 7 TeV.The charged particles get this energy by radio frequency radiation from super-conducting cavities operating at a temperature of 4.5 K (= −268.7 ◦C) and at afrequency of 400 MHz. For each beam there are eight cavities, each delivering anaccelerating field of 5 MV/m. To keep the particles circulating around the ringand inside the beam pipe (vacuum pressure of 10−13 atm), a magnetic field of8.3 T is used. This field is generated by superconducting dipole electromagnetsoperating at a current of 11.7 KA and at a temperature of 1.9 K (= −271.3 ◦C).Superfluid helium acts as cooling fluid. It has a very high thermal conductivity,

18 Chapter 3. LHC and ATLAS

Figure 3.1: Schematic view of the LHC-accelerator and its four experiments. [31]

which is intended to stabilize the large superconducting system. Figure 3.2 showsthe cross section of a LHC dipole. In total, there are 1232 dipoles of 15 m lengthand 35 t weight in the LHC ring cooled by 120 t of helium.

Without further magnetic fields the charged particles repelling themselveswould deviate from the ideal trajectory, diverge in the plane transverse to thebeam and be absorbed by the beam pipe vacuum vessel. To keep the particlesdensely together, many other focussing magnets, including 392 quadrupoles, areused. Such magnets are also used around the interaction points, in order tosqueeze the particles as much as possible to increase the probability of a collision.

Due to the radio frequency acceleration scheme, the protons circulating in-side one beam pipe ring are divided into bunches of about 1011 particles. 2808of these bunches circumnavigate the ring at nearly the speed of light (v/c =0.999 999 991). Thus, they traverse the entire ring about 10 000 times per second.The bunches follow each other at separation of about 7.5 m in length or 25 ns intime. This corresponds to a collision rate of ≈ 40 MHz at the interaction points.

The particle current density in the LHC is specified by the machine parameterL, named luminosity. As defined in equation 3.1, the (instantaneous) luminosity

3.1 The Large Hadron Collider (LHC) 19

Figure 3.2: Cross section of the vacuum pipes inside a dipole magnet. [32]

L is the product of the numbers of particles n1, n2 in both crossing bunches andthe frequency f of bunch crossings, divided by the cross sectional area A of abunch:

L =n1 · n2 · f

A(3.1)

The full formula according to the LHC design (beam profile) can be found in [33].The time between beam commissioning at injection energy of 450 GeV per

beam and the first physics run at design center of mass energy of 14 TeV anddesign luminosity of L = 1034 1

cm2·s will be probably more than one year. Atthis luminosity, there are about 23 inelastic scatterings per bunch crossing, inwhich nearly 1000 new particles are produced. The rate R of the proton-protonreactions inside the LHC machine is given by the product of the proton-protoncross section σ and the luminosity: R = σ · L.

In the next section, we present the LHC experiment ATLAS in detail, wheresuch particle collision events are detected and analyzed.


3.2 The ATLAS Experiment

ATLAS is a multi purpose detector [34], that is designed to observe particlecollision events within the complete LHC parameter range. ATLAS is about44 m long, has a diameter of approximately 25 m and a weight of about 7000 t.Due to the design of the LHC as a ring collider, ATLAS has a nearly cylindricallysymmetric structure (barrel with 2 endcaps). Figure 3.3 shows a detailed view ofATLAS.

Figure 3.3: Detailed view [35] of the ATLAS detector and its subdetectors

To identify positions within ATLAS, the following coordinate system is used: Thebeam direction defines the z-axis, and the x-y plane is the plane transverse to thebeam direction. The positive x-axis is defined as pointing from the interactionpoint to the center of the LHC ring, and the positive y-axis is pointing up-wards. [26] The ATLAS side that is in positive z-direction is called A-side, theother side is called C-side. The azimuthal angle φ is measured around the beamaxis, and the polar angle θ is the angle from the beam axis. The pseudorapidity isdefined as η = − ln (tan(θ/2)). The distance ∆R in the pseudorapitidy-azimuthalangle space is defined as ∆R =

√

∆2η + ∆2φ. [26]In a LHC collision event many particles are produced. Some types of particles canbe directly detected. But since particles differ in their properties and interactions,ATLAS is a combination of different specialized particle detectors: The inner

3.2 The ATLAS Experiment 21

tracking detector [36], the electromagnetic and hadronic calorimeters [37], and themuon spectrometer [38], that is the outermost detector. The spatial arrangementof the subdetectors and an event cross section is sketched in figure 3.4. The

Figure 3.4: Schematical view of the specialized particle detectors of ATLAS and sketch of anevent cross section. Not drawn: the toroid magnets between the hadronic calorimeter and themuon spectrometer; the radiation shields. [39]

subdetectors are explained in detail in the following sections. By measuringprecisely the directly detectable particles, other particles can be reconstructed:For example, some particles exist after production at the interaction point forsuch a short time that they do not leave the beam pipe before they decay.

The momenta of charged particles can be measured by applying a magneticfield and reconstructing the trajectories of the particles. In ATLAS, there arefour superconducting magnets [40]: one solenoid [41] with a nominal strengthof 2 T for the inner detector, and three air-core toroids (one barrel [42], twoendcaps [43]) with a peak strength of about 4 T for the muon spectrometer.

Due to the high design luminosity and energy of the LHC beams, a lot ofsecondary radiation is generated in the detector. To reduce this background


radiation, nearly 3000 t of shielding material [44] has been added to the detector.For the operation and control of the ATLAS detector following systems are needed[45, 46]:

• CERN infrastructure systems including safety, cryogenics, cooling and ven-tilation, gas, electricity, vacuum

• CERN magnet system

• LHC conditions and beam interlock system (BIS) [47], LHC fast signals [48]and timing, trigger, control system (TTC) [49]

• Finding Persons Inside ATLAS Area system (FPIAA) [50]

• Detector Safety System (DSS) [51]

• Detector Control System (DCS) [52]

• TDAQ: Trigger and Data AcQuisition (DAQ) [53, 54]

The control and data flow of the ATLAS experiment is illustrated in figure 3.5.All communications between the CERN infrastructure, the magnets, the LHC

Figure 3.5: ATLAS control and data flow: DCS, (T)DAQ and external services. Dotted linesindicate DCS connections, dashed lines represent timing signals, dashed-dotted lines indicateinformation from or to external services, solid lines represent trigger or data information. Allacronyms are explained in the text.

machine, the FPIAA, the DSS and the TDAQ are done via DCS, with the ex-ception of fast signals (such as the LHC 40.079 MHz bunch crossing clock and11.246 KHz orbit signal) that are distributed by the TTC system and handled inthe level-1 (LVL1) trigger [55]. The DCS has to work continuously to ensure the


safe operation of the detector. The DCS guarantees also a reliable coordinationwith the LHC control system and with other external services. Even when TDAQis not available, DCS controls hierarchically all subdetectors, common servicesand the environment in the cavern.

The DCS is divided into the Front-End (FE) and the Back-End (BE). TheBE is organized in three layers, the Global Control Stations (GCS) with serverapplications and human interfaces in the ATLAS control room for the overalloperation, the Sub-detector Control Stations (SCS) for high-level control of asubdetector or of common services (e.g., cooling or TDAQ devices), and the LocalControl Stations (LCS) for process control of subsystems. The LCS controls alldevices in the cavern via the FE. The FE is based on Controller Area Network(CAN) field buses deployed over the whole experiment and uses Embedded LocalMonitor Boards (ELMB), that are insensitive to the magnetic fields and tolerantto radiation. The DCS stores all ATLAS detector conditions data in a relationaldatabase, which is important for the offline reconstruction and analysis.As illustrated in figure 3.6, the LHC beam has to be stable and the DCS must

Figure 3.6: ATLAS can be operated, if LHC, DCS and DAQ have status safe and ready. [53]

have enabled the detector and the TDAQ, so that data can be taken.

The TTC system synchronizes ATLAS and LHC, so that the hit data of allsubdetectors can be assigned to the correct bunch crossing. The default bunchcrossing rate is ≈ 40 MHz, and at design luminosity every candidate event for newphysics will be accompanied by 23 inelastic events on average. However, the eventdata recording is limited to about 200 Hz because of technology and resource lim-itations. Therefore, an overall rejection factor of 5 · 106 against minimum-biasprocesses is required while maintaining maximum efficiency for the new physics.The selection (trigger) of events and their transfer to the storage (DAQ) is doneby the TDAQ.


The trigger system consists of three levels of event selection: level-1 (LVL1) trig-ger, level-2 (LVL2) trigger and event filter (EF). The LVL2 and EF together formthe High-Level Trigger (HLT). The LVL1 trigger is implemented using custom-made electronics, while the HLT is almost entirely based on commercially avail-able computers and networking hardware.The LVL1 trigger uses a subset of the total detector information (parts of thecalorimeter and of the muon spectrometer) to make a decision on whether ornot to continue processing an event. The LVL1 trigger reduces the data rate toapproximately 75 KHz (limited by the bandwidth of the readout system, whichis upgradeable to 100 KHz). If an event is accepted, a LVL1 accept signal (L1A)is fed into the TTC system. When the L1A signal reaches the Read Out Drivers(ROD) of the subdetectors, all hit data of the requested event are read out andtransferred to the Read Out Buffers (ROB). Furthermore, the LVL1 trigger iden-tifies regions in the detector where it has found interesting features, so-calledRegions of Interest (RoI). The RoI information from the various parts of theLVL1 trigger are combined by the RoI Builder (RoIB) and then passed to theLVL2 trigger.For those events fulfilling the LVL2 selection criteria (maximal acceptance rate:3.5 KHz), event-building (EB) is performed. The assembled events are thenmoved by the DAQ to the EF, and the events selected there are moved topermanent event storage (200 Hz) at CERN. Each event has a total size ofabout 1.5 Mbytes and is stored in a Raw Data Object file (RDO). Hence, severalpetabytes of raw data will be stored at CERN per year.

This data is analyzed [56, 57] via the LHC Computing Grid (LCG) [58, 59].

Figure 3.7: Calibration and physics analysis in the LCG-GRID.

According to a four-tiered model [60] illustrated in figure 3.7, the data is dis-tributed around the globe. A primary backup will be recorded on tape at CERN,


the Tier-0 center of the LCG. After primary event processing at the Tier-0, theraw and the primary processed data is copied to a series of Tier-1 centers. Theselarge computer centers have sufficient storage and reprocessing capacity withround-the-clock support for the Grid. At the Tier-1 centers, derived datasetsare produced as Event Summary Data files (ESD) or Analysis Object Data files(AOD). For analysis, this data is distributed to the Tier-2 facilities, that are ac-cessed by individual scientists, who use Tier-3 facilities, such as a workstation.For quick event access and histogramming, Derived Physics Data files (DPD)can be produced out of the AODs at the Tier-2 centers and stored at the Tier-3facilities.

For calibration and analysis an object-oriented approach to software is used,based on the programming language C++ [61, 62]. Other languages are alsoused, including FORTRAN [63], Java [64] and Python [65]. The Athena frame-work [66], that has a component-based architecture, allows one to perform allneeded activities such as: calibration and alignment, simulation (event gener-ation, detector simulation, digitization, pile-up) and analysis (event selection,reconstruction, physics analysis).

3.2.1 The Muon System

Final state muons are very important in many analyses and provide robust signa-tures of physics at the LHC, for example the Higgs boson decay H → ZZ∗ → 4µor a semileptonic channel of a decaying top/antitop quark pair tt → 4j + /ET +µ.Therefore, the muon spectrometer has a high resolution and a stand-alone trigger-ing momentum measurement capability. Since muons penetrate the large amountof material where all other particles are absorbed, in order to measure their en-ergy, the muon spectrometer [38] is the outermost subdetector of ATLAS. Asexplained above, the muon spectrometer measures the charge and momentum ofmuons by reconstructing their trajectories in the magnetic field of the air-coretoroids. The barrel chambers form three cylinders concentric with the beam axisand cover the pseudorapidity range |η| < 1. The outer barrel chambers are 10 maway from the beam pipe, their half length is 12.5 m. The forward chambers coverthe range 1 < |η| < 2.7 and the outermost are about 21.5 m from the interactionpoint. In order to provide a resolution of σ(pT )/pT = 10% at pT = 1 TeV, thespace resolution must be at least 50 µm in z and 0.5 mrad in φ.

Several types of muon tracking chambers exist: Monitored Drift-Tube cham-bers (MDT) and Cathode Strip Chambers (CSC) are precision chambers, Re-sistive Plate Chambers (RPC) and Thin Gap Chambers (TGC) are fast LVL1trigger chambers. The principle of measurement is the same for all four types:Muons, that cross a gas gap between an anode and a cathode (for example awire inside a tube or two parallel plates), cause a local gas discharge and hencean electric signal, that can be read out. In total, there are 5376 chambers or1.0757 · 106 readout channels.


The MDTs provide a precision measurement (mechanical accuracy of 30 µm) ofthe track coordinates (80 µm single wire resolution) over most of the η-range. Atlarge pseudorapidities and close to the interaction point, CSCs are used, thathave a higher granularity and withstand the demanding rate and backgroundconditions.The trigger chambers have a time resolution better than the LHC bunch spacingof 25 ns to identify the bunch crossing. The requirements on the time and spaceresolution are met by following specialization: The RPCs have the better timeresolution (1.5 ns), whereas the TGCs have the better spatial resolution (anodewire pitch 1.8 mm).

Over the large global dimensions of the muon spectrometer, it is not pos-sible to stabilize the dimensions and positions of the chambers at the required30 µm level. Therefore, chamber deformations and positions are constantly moni-tored by means of optical alignment systems. As a result, displacements between≈ 30 µm and ≈ 1 cm can be corrected for in the offline analysis [26].

3.2.2 The Calorimeters

The ATLAS calorimeters [37] are crucial in many measurements, for examplefor the Higgs boson decay H → γγ or the semileptonic electron channel of adecaying top/antitop quark pair tt → e + /ET + 4j. In order to identify events,the calorimeters must separate well electrons/positrons, photons and jets. Whenelectrons/positrons or photons enter the calorimeter, an electromagnetic showeris initiated. The shower breaks down and is absorbed, when secondary photonsfall below the pair production threshold due to energy losses. Whereas hadrons,which are the main constituents of jets, rather initiate hadronic showers thanelectromagnetic ones. Since both types of showers have different characteristics,for example electromagnetic showers begin a shorter distance into the calorimeterthan hadron showers, the origin of a shower may be determined.Hence, the calorimeters consist of an outer hadronic calorimeter (HCAL) andan inner electromagnetic calorimeter (ECAL), both covering the pseudorapidityregion |η| < 3.2. A dedicated dense liquid argon forward calorimeter (FCAL)with rod shaped electrodes in a tungsten matrix covers the range 3.1 < |η| < 4.9,where the level of radiation is very high. The total weight of the calorimetersystem including the solenoid flux return iron yoke is nearly 4000 t.

Since the calorimeters contribute to the selection of events at the LVL1 trigger,their corresponding parts have a fast response time of few nanoseconds. TheECAL is dedicated to the measurement of energy and position of electrons andphotons (and separation of γ/π0, e/π, etc.). Whereas, the HCAL measures theenergy and position of jets and separates hadronic decays of τ -leptons from jets.Both calorimeters, especially the HCAL, contribute to the reconstruction of themissing transverse momentum of the event. The region 1.37 < |η| < 1.52 is notused for precision physics measurements involving photons because of the large


amount of material (about 7 radiation lengths) situated in front of the ECAL(cryostats).

The HCAL is a scintillation calorimeter. Its barrel consists of plastic scin-tillator plates (tiles), in which light flashes are produced, when high energeticcharged particles (or photons) traverse the plates. These are embedded in aniron absorber and read out by photomultipliers. This tile calorimeter system [67]has an outer radius of 4.25 m and a total length spanning of ±6.10 m. At largerrapidities, where higher radiation resistence is needed, the intrinsically radiationhard liquid argon technology is used. Due to the high granularity (∆η · ∆φ) be-tween 0.1 · 0.1 and 0.2 · 0.2 (representing about 104 readout channels) the HCALprovides a precise measurement of the three-dimensional jet direction (from thedirection of the shower) and a good jet resolution. By also measuring the showerenergy with the calibrated photomultipliers the four-momentum of a jet maybe obtained. In the range |η| < 3, the energy resolution is ∆E

E= 50%√

E⊕ 3%.

Moreover, the jet energy scale is linear within 2% up to an energy of 4 TeV. Inorder to absorb all particles except muons in front of the muon spectrometer, theHCAL has a thickness of about 10 interaction lengths. This leads to the so calledshower containment inside the HCAL and to a reduction of background in themuon chambers.

The ECAL is a lead/liquid argon (LAr) detector with accordion geometry.Together with the hadronic endcap calorimeter (HEC), and the FCAL it formsthe Liquid Argon Calorimeter [68], which are types of ionization calorimeters.The whole LAr calorimeter has a length of ±6.65 m along beam axis and anouter radius of 2.25 m, giving a total thickness of about 25 radiation lengths.When high energetic charged particles (or photons) traverse the liquid argon,argon atoms are ionized and free electrons are produced. In an electrical field,that is generated by an applied high voltage, the electrons drift to the cathodeand induce a current there, which is read out. By an appropriate calibration theshower energy and hence the energy of the high energetic particle is measured.In order to identify (electron/photon/tau – jet separation) and resolve particleswell, the granularity (∆η ·∆φ) of the ECAL is between 0.003 ·0.1 and 0.05 ·0.025,resulting in a total number of readout channels of about 1.9 · 105. The ECALhas an energy resolution of about 10 %/

√

E[GeV] and can reconstruct electronsfrom 1 GeV to 5 TeV. The energy scale precision is better than 0.1%, and upto 300 GeV the linearity of response is better than 0.5%. Within 10 years ofoperation the ECAL has to withstand neutron fluences of up to 1015 n/cm2 andradiation doses of up to 2 · 105 Gy.

The amount of material in front of the muon system, that includes the supportstructure of the Tile calorimeter, is about 11 interaction lengths. Whereas, thetotal material (Inner Detector and services) seen by an incident particle up tothe ECAL front face is circa 2 radiation lengths.


3.2.3 The Inner Detector

The Inner Detector (ID) [36] is very important for physics analyses, because itis the main responsible for reconstructing tracks, (primary) vertices and trackimpact parameters as well as for electron identification (between 0.5 GeV and150 GeV). Especially its contribution to τ - and heavy flavor (c-, b-) tagging isessential for precision measurements such as of the top quark mass, as we will seein Chapter 6. The inner detector is a tracking detector with a length of 7 m anda radius of 1.15 m, a quarter section of it is illustrated in figure 3.8. The ID has

Envelopes

Pixel

SCT barrel

SCT end-cap

TRT barrel

TRT end-cap

255<R<549mm|Z|<805mm

251<R<610mm810<|Z|<2797mm

554<R<1082mm|Z|<780mm

617<R<1106mm827<|Z|<2744mm

45.5<R<242mm|Z|<3092mm

Cryostat

PPF1

CryostatSolenoid coil

z(mm)

Beam-pipe

Pixelsupport tubeSCT (end-cap)

TRT(end-cap)

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8

Pixel

400.5495

580

650749

853.8934

1091.5

1299.9

1399.7

1771.4 2115.2 2505 2720.200

R50.5R88.5

R122.5

R299

R371

R443R514R563

R1066

R1150

R229

R560

R438.8R408

R337.6R275

R644

R1004

2710848712 PPB1

Radius(mm)

TRT(barrel)

SCT(barrel)Pixel PP1

3512ID end-plate

Pixel

400.5 495 580 6500

0

R50.5

R88.5

R122.5

R88.8

R149.6

R34.3

Figure 3.8: Plan view of a quarter-section of the ATLAS inner detector showing each of themajor detector elements with its active dimensions and envelopes. The labels PP1, PPB1 andPPF1 indicate the patch-panels for the ID services. [34]

a barrel-disk structure like the other detectors and it measures the momentumof tracks in the pseudorapidity range |η| < 2.5.The ID consists of three subdetectors: the Transition Radiation Tracker (TRT)[69], the SemiConductor Tracker (SCT) [70] and the Pixel detector [71]. Due tothe size and arrangement of the barrel layers and disks of the pixel and SCT de-tector, typically seven space points are provided by the semiconductor detectors.The TRT measures usually 36 tracking points. The resolutions of the subdetec-tors are given in table 3.1.


Similar to the muon chambers the TRT uses straw tubes (aluminum coatedpolyamid) with a length of 1.44 m (barrel tubes) and a diameter of 4 mm. Dueto this small diameter and the isolation of the sense wires (30 µm diameter,gold-plated tungsten-rhenium) within the individual gas volumes, the TRT canoperate at the very high rates expected at the LHC. By employing xenon gasit is also possible to detect transition radiation photons created in a radiator(polypropylene-polyethylene) between the straws. This technique is intrinsicallyradiation hard and allows the identification of electrons as well as a large numberof measurements, typically 36, on every track. The resolution per straw is ≈130 µm. The TRT consists of 96 modules in the barrel and 20 modules per endcap, comprising about 3.51 · 105 readout channels in total.

Subdetector σR−φ [µm] σz [µm] readout channels [106]Pixel 10 115 80.4SCT 17 580 6.3TRT 130 per straw 0.351

Table 3.1: Key numbers of the ATLAS inner detector.

The SCT detector is made up of 4 barrel layers and 9 disk layers per end cap.In total, there are 4088 modules consisting of about 6.3·106 readout channels. TheSCT detector contains about 63 m2 of small (6.36 · 6.40 cm2) silicon microstripdetectors, that consist of 768 readout strips of ≈ 80 µm pitch each. When acharged particle traverses a strip, free electrons and holes in a low doped fullydepleted silicon wafer are produced. Some electrons and holes are collected atelectrodes and induce there the readout current.Since always two small microstrip detectors are arranged back-to-back, the SCTprovides on average 8 precision points in the Rφ and z coordinates (4 spacepoints). In order to obtain the z measurement, small angle (40 mrad) stereobetween both small detecors is used. Tracks can be distinguished, if they areseparated by more than 200 µm. The resolution in Rφ is 17 µm, in z 580 µm.Like the pixel detector, the SCT requires very high dimensional stability andcold operation. By cooling, the radiation damages are reduced, that will alterthe fundamental characteristics of the silicon wafers.

The Pixel detector is the innermost subdetector of ATLAS and thereforedetermines mainly the impact parameters of tracks. This detector consists of 3barrel layers and 3 disk layers per end cap, comprising 1744 modules or about8 · 107 readout channels in total. The Rφ resolution is 10 µm, the z resolution is115 µm.Since a part of this thesis was the participation in and support of production,installation and commissioning of optoelectronics for this detector, more detailsabout it and about that work are provided in the next chapter.

Chapter 4

Optoelectronics of the PixelDetector System

4.1 The Pixel Detector

As already mentioned in Section 3.2.3, the pixel detector is the innermost sub-detector of ATLAS. The innermost layer of the pixel detector is 5 cm away fromthe beam pipe. Due to this proximity and its very good point resolution ofσr−φ < 15 µm and σz < 120 µm, mainly the pixel detector determines the trackimpact parameter. Furthermore, the pixel detector is the main responsible forthe resolution of primary and secondary vertices. The determination of track im-pact parameters and vertices is very important for many physics analyses. Thisis especially true for top quark pair events, if b-tagging is used. More detailsabout this and a procedure for the calibration of track impact parameters will bepresented in Chapter 6.

The pixel detector consists of 3 barrel layers and 3 disks on each forward side,as illustrated in figure 4.1. By this geometry, the detector acceptance is withinthe pseudorapidity range of |η| < 2.5. As a consequence of the proximity to thebeam pipe, the pixel detector must be so radiation hard that even at the end ofits lifetime it delivers 3 spacepoints for each traversed charged particle. Moreover,there are three further requirements of the pixel detector: First, the readout hasto be at least as fast as the level-1 trigger frequency of 75 KHz. Second, pixel hitsmust be correctly assigned to the corresponding bunch crossing (bunch crossingrate: ≈ 40 MHz). Third, in order to minimize the error of measurement of thecalorimeters, radiation and interaction length of the pixel detector have to beminimal.

The barrel layers consist of staves that have a width of 16 mm and a lengthof 830 mm. In order to minimize the interaction and radiation length of the pixeldetector, the support structures of the staves are made of carbon fiber. On thefar side (seen from the beam pipe) of a stave, there is a cooling pipe so that the

32 Chapter 4. Optoelectronics of the Pixel Detector System

Figure 4.1: A schematic view of the pixel detector in its mounting structure. [72]

electronics is kept at the operation temperature of −10 ◦C and radiation damageis minimized. On the near side, there are the active elements, the pixel modules.On each stave, there are 13 modules, each with 47232 pixels as sensors [73, 74].For reasons of space, there are 4 ganged pixels in each column of the front-endchip, thus leading to a total of 46080 readout channels. About 90% of all pixelshave an area of 50 µm× 400 µm each, the other have an area of 50 µm× 600 µmeach.

On the other hand, a disk, that has an inner radius of 89 mm and an outerradius of 150 mm, consists of 8 sectors. Each sector holds 6 modules (3 on thefront, 3 on the back), which are identical to those on the staves. Table 4.1 provides

Radius (mm) Staves ModulesLayer-0 50.5 22 286Layer-1 88.5 38 494Layer-2 122.5 52 676

z-Position (mm) Sectors ModulesDisk-1 ±495 8/8 48/48Disk-2 ±580 8/8 48/48Disk-3 ±650 8/8 48/48Total 1744

Table 4.1: Dimensions and coordinates of the ATLAS pixel detector subparts.

the dimensions and coordinates for the subparts of the pixel detector. In total,there are 1744 modules or 80 363 520 readout channels, which represent more than

4.1 The Pixel Detector 33

90% of all readout channels of ATLAS. As a consequence, it is very vital that thepixel detector is carefully and successfully operated as well as calibrated.

Figure 4.2 shows the structure of a pixel module. Each pixel is connected toa single electronic cell via a tiny solder bump (bump bond). When a chargedparticle traverses a pixel, free electrons and holes in the n+ on n type sensorare produced. Over the bump bond, a capacitor is charged, which delivers thereadout current.

Figure 4.2: At the left: scheme of a pixel module. The high voltage (HV) for the sensordepletion is led through a hole in the flex to the pixels. A guard ring prevents sparkovers. Forthe thermal contact to the cooling pipe the pixel modules are glued onto the carbon-carbonthermal management tiles (TMT). More information can be gained from the text. At the right:top view photo of a pixel module. The top side points to the beam pipe. [34]

2880 electronic cells form one special ASIC, the Front-End (FE) chip [75]. Oneach module there are 16 FE chips, that amplify and digitize the signals from thesensors. The FE chips are connected via 25 µm thick wirebonds, held by captonfoil, to the flexible copper-capton board, the “flex”.Several components are on the flex, including a negative temperature coefficient(NTC) temperature sensor, decoupling capacitors and the Module Control Chip(MCC) [76]. The MCC configures the 16 FE chips, sends to them received triggersignals and reads their hit data. These electrical trigger signals and hit data cometo or leave the module via the “pigtail” and its Type0 connector. The electricalsignals come from or go to the optoboard (OB). More about the readout of thepixel modules will be explained in the next section.


DataOut1, DataOut2

~80 m Fibre~1 m cable

TX−plugins

RX−plugins

S−LINK

(BPM)

(DRX)

(PiN, DORIC,

Clock, DataIn,

(MCC + 16 FE chips)

Module Optoboard

VDC, VCSEL) Data & Commands

Eventdata

Module data

Control

BOC ROD

Read Out Buffer

Builder

Event

TTC link

Data link

Electrical communication

Clock TIM

from LHC

machine

TTC−signal

Control

Off−detectorOn−detector

Crate internal communication

read outby LV2 trigger

S−Link optical transmission

Communication with main detector machine

Pixel specific optical transmission

Readout crate

��

��

��

��

��

��

��

Figure 4.3: The ATLAS pixel dectector readout chain. [11]

4.2 The Pixel Detector Readout: The Optical

Link

In figure 4.3, the readout chain of the ATLAS pixel detector is shown. It consistsof two parts: the on-detector part, that is located inside the ATLAS detector inthe cavern UX15, and the off-detector part, which is located outside the ATLASdetector in the cavern USA15. Due to the high data rate and for an electricalseparation of both parts, the on- and off-detector sides are connected by fiberoptical cables. Therefore, on each side an opto-electrical interface is needed.

As indicated in the previous Section 4.1, on the on-detector side the electricalsignals from a module are converted by the optoboard (OB) [77] to optical signalsand sent to the off-detector side. There, the Back Of Crate (BOC) card [78]receives the optical signals and converts them to electrical signals. Vice versa theOB converts optical signals coming from the BOC card to electrical signals andsends them to the MCCs. 1 BOC card is connected by long glass fiber ribbons toup to 4 OBs. On the BOC side, each ribbon consists of 8 fibers (on the OB side,two ribbons of 8 fibers each form one ribbon of 16 fibers). Each fiber representsthe connection to 1 MCC. The data transmission is done unidirectional. Thismeans that for each MCC 2 fibers (3 for a Layer-0 MCC) are necessary, one forsending commands to the MCC and one (two for a Layer-0 MCC) for receivingdata from the MCC. All 8 fibers of 1 ribbon are either for sending data to or forreceiving data from MCCs.

Each BOC card is connected back to back to a Read Out Driver (ROD) [79]in a 9U VersaModule Eurocard (VME) crate. As the name indicates, the Backof Crate card is located on the back side of the ATLAS pixel readout crates.

4.3 Production, Installation and Commissioning of Optoelectronics 35

Each crate can hold up to 16 RODs/BOC cards and contains 1 Timing, Triggerand Control (TTC) Interface Module (TIM) [80] and 1 Readout Crate Controller(RCC). This is a Single Board Computer (SBC) and steers the cards in the crate.The TIM receives the optical TTC signal. It sends e.g. a Level-1 Trigger “accept”signal (75 KHz) electronically to the RODs. The TIM sends also a clock signal tothe BOC cards. The signal has a frequency of ≈ 40 MHz, which corresponds tothe bunch crossing rate, and it is tuned on the BOC card to compensate signalpropagation delays.

The BOC card receives the readout commands from the ROD and encodesthem with the clock into one Bi-Phase Mark (BPM) [77] signal. This encodingreduces the number of needed fibers and opto-transmitters/receivers. The electricBPM signal is converted to an optical signal and sent to the OB. The OB receivesthe optical signal and converts it back to an electrical signal. Then, the BPMsignal is decoded, and clock and command data are restored. Both are sentseparately from the OB to the modules.

Likewise, the OB receives the data from the MCCs and converts it to opticalsignals. The BOC card receives these signals, converts them to electrical dataand sends it to the ROD. Here, the data format is checked, and a reformatting ofthe data is performed. Thereafter, the data is processed by the Event FragmentBuilder. Afterwards, the event fragment data is sent to the BOC card, thatconverts the electronic signals to optical signals and sends them to the Read OutBuffer (ROB). This buffers event data with a frequency of 75 KHz until a Level-2Trigger “accept” signal arrives (2 KHz). The further steps in the trigger and dataacquisition chain were explained above in Section 3.2.

More details about the optical link can be found in [81]. The BOC card andthe optical cables, their production, installation and commissioning, are describedin the following section.

4.3 Production, Installation and Commission-

ing of Optoelectronics

4.3.1 Back Of Crate Cards

In the previous section, the BOC card, its position in the readout chain of theATLAS pixel detector and its main tasks were introduced. This section summa-rizes the production, installation and commissioning of the BOC cards in detail.The main responsibility for these tasks was taken by the Wuppertal group.

In total, 132 BOC cards are needed for the ATLAS pixel detector. The cardsare placed in 9 readout crates in the 5 racks Y25-05A2 to Y29-05A2 in the ATLAScounting room USA15.The BOC card is a 9U VME 64X transition card, which was developed by thegroup of the Cavendish Laboratory in Cambridge [82]. The card was designed


for the SCT and the pixel detector, and the board design is the same for both.However, a section on the BOC card is unused and not assembled, if it is employedfor the pixel detector. This is because of the different detector design. Figure 4.4shows a photo of a pixel BOC card.

Figure 4.4: A Back Of Crate Card card in pixel assembly (Layer-2). [11]

Each pixel BOC card can serve up to 32 pixel modules (channels), but due tomodularities not all channels of a BOC card will be used. Furthermore, thenumber of used channels depends on which part of the pixel detector the BOCcard has to support. For Layer-0, one BOC card is connected to one OB by 3optical 8-way ribbons (these are described in Section 4.3.2), 1 for the transmissionof clock and commands to the OB, 2 for the transmission of data to the BOCcard. For Layer-1 and Disks, one BOC card is connected to 2 OBs by 2 opticalribbons each, 1 for clock/commands, 1 for data. For Layer-2, one BOC card isconnected to 4 OBs by 2 optical ribbons each, 1 for clock/commands, 1 for data.The BOC card can be operated in 40 or 80 MBit/s mode, determining the datatransfer rate from the OB to the BOC card. For Layer-2 this rate will be 40MBit/s, for Layer-1 and Disk-1,2,3 it will be 80 MBit/s. The 160 MBit/s ratefor the Layer-0 is achieved by two connections each operated at 80 MBit/s. Thenumbers of required BOC cards for each configuration are shown in table 4.2.

The BOC card has to provide the following functionalities:

• Receive electrical commands from the ROD

• Receive electrical clock from the TIM

• Convert electrical commands and clock to Bi-Phase Mark encoded opticalsignals and send them to the OB(s); provide masking, timing and lasercurrent adjustment functions for these signals


Detector part OBs BOC cards bit data rate

Layer-0 44 44 160 MBit/sLayer-1/Disk-1,2,3 76 + (16+16+16) 38 + (8+8+8) 80 MBit/s

Layer-2 104 26 40 Mbit/s

total 272 132

Table 4.2: Number of BOC cards to read out the complete ATLAS pixel detector, split upfor the different detector parts.

• Receive raw optical data from the OB(s) and convert them to raw electricaldata; provide a timing and threshold adjustment and clock synchronizationfor these signals

• Send raw electrical data to the ROD

• Receive electrical event fragment data from the ROD

• Convert electrical event fragment data to optical data and send them viathe S-Link path to the ROB

• React to the laser safety interlock

The layout of the BOC card and those tasks, which represent a signal flow on theBOC card, are sketched in figure 4.5. Several functionalities are separated intoseveral sections on the board: the 2 VME bus connectors (1 for input signals,1 for output signals) to the backplane; the main Complex Programmable LogicDevice (CPLD), that controls the board; the clock section, where the timing canbe tuned; the receive section with the mounted RX-plugins, which convert theincoming optical signals to electrical signals via PIN diodes (P-type, Intrinsic,N-type); the command and transmission section with the mounted TX-plugins,that encode the clock and the commands from the ROD with the BPM-chip [83]into one signal, which is converted by the Vertical Cavity Surface Emitting Laser(VCSEL) [83] arrays into the outgoing optical signals; the S-Link [84,85] sectionwith the mounted S-Link card, or High speed Optical Link for ATLAS (HOLA);the 3 voltage lines (not shown in the figure), one with 3.3 V, one with 5.0 V andone with 12.0 V; the 2 hexadecimal rotari switches (not shown in the figure), thatdefine the serial number of the BOC card. More information about the designand the functionality of the BOC card can be found in [11].

The main assembly of the BOC circuit boards was done by the Turck company[86] in Germany. It delivered 156 BOC circuit boards, leaving 24 as spare boards.The postproduction at the University of Wuppertal consisted of 6 steps for everyBOC circuit board:


Figure 4.5: Sketch of the signal flow on the Back Of Crate card (Layer-2 assembly arbitrarilychosen). [87]

1. Optical Inspection, Assembly

2. Electrical Power Test/Programming

3. Main Electrical Test (E-Test I)

4. Mechanical Assembly

5. Optical Test (O-Test), Assembly

6. “S-Link” Test (E-Test II)

The numbers of required components for the postproduction in Wuppertal canbe found in table 4.3. In step 1, the delivered board was inspected optically. Ifa problem was found, the board was sent back to the Turck company for repair.Common problems were: The packing was improper so that a board was slightlydamaged (e.g. scratches) during transport. Components were shifted or rotated.There were missing or excess components. Some components were defect. Oftenthere were shorts at adjacent soldering joints. Sometimes, cold soldering joints


component total number number ofof pieces pieces needed for ATLAS

BOC Circuit Boards 156 132Front-Panels 156 132

Routing Boards 156 132S-Link-Cards 156 132

TX-Housings/RX-Housings 156/312 132/264TX-Plugins/RX-Plugins 315/366 272/316

Washers/Screws ≈ 2000/6000 ≈ 1700/5100

Table 4.3: Numbers of BOC card components.

were found. If the board was accepted, it was serially numbered, and a smallrouting board and handles were assembled.

In step 2, an electrical power test was done. Here, the voltages and currentsof the 3 power lines of the board are measured. The corresponding currents andtheir allowed ranges depend on the assembly and are shown in tabular 4.4. If thevoltages or currents were not in the requested ranges, the board was investigatedfurther (mostly there were shorts at soldering joints).

current range 3.3 V line 5.0 V line 12.0 V line

main assembly 1.8–2.2 A 0.45–0.6 A 5–10 mAcomplete assembly −2.8 A −0.9 A −15 mA

Table 4.4: The current ranges for the three voltage lines of the BOC card. The ranges dependon the assembly stage of the BOC card.

Next, the 7 Complex Programmable Logic Devices (CPLDs) of the BOC cardwere programmed (firmware). If this failed, the board was investigated further.

In step 3, the main electrical test (E-Test I) was done. The 5 LEDs of theBOC card were tested: two green LEDs for the power lines U=5.0 V and U=3.3 V,two yellow LEDs for the external (Patch Panel 1, PP1) and internal (crate door)interlock, one red LED for lasers on (no interlock). With test plugins the TX-sites were connected electrically with the RX-sites (“loopback”). In a later step,the plugins for transmitting/receiving light to/from the OBs are mounted here.With the test plugins the following was tested: the data paths (100% successrequired), the control of the plugins (100% success required), the timing (clocks),some other functionalities (for example the PIN diode current limitation). De-tails of the electrical test procedure can be found in [11]. If the electrical test wasnot 100% successfully, further investigation of the board was carried out.


In step 4, the S-Link-Card, the RX/TX-Housings and the RX/TX-Plugins,which were produced in Taiwan and are illustrated in figure 4.6, were mountedon the BOC board.

Figure 4.6: Left: a TX-Plugin. Right: an RX-Plugin. [11]

In step 5, the optical test was done. The TX-plugins were connected to theRX-plugins by optical ribbons (“loopback”). So, the mechanical alignment ofthe plugins/housings was verified, if light from the TX-plugins arrived for all 8channels each at the RX-plugins. Furthermore, the TX/RX-chips could be tested.Details of the optical test procedure can be found in [11]. If the optical test wasnot passed, the plugins were realigned or changed. After passing the optical test,the front panel was assembled.

In step 6, the “S-Link” test (E-Test II) was done. During this, the remaininguntested components of the BOC card were tested electrically by sending datafrom the ROD via the S-Link card of the BOC card to the ROB.

In table 4.5, the current status of the BOC card production is shown. Analways up to date status can be found in [88, 89]. At the moment, there are 143BOC cards ready for operation, but only 132 are needed for the ATLAS pixel

Table 4.5: Current status of BOC card production.

Step Needed for ATLAS Done

Main Assembly 132 156Incoming Inspection 132 153Electrical Power Test/Programming 132 153Main Electrical Test (E-Test I) 132 153Mechanical Assembly 132 143Optical Test 132 143“S-Link” Test (E-Test II) 132 143

completely finished 132 143


detector. So, there are 11 spares (8%), that can be used for testing purposes.If there would be more TX-plugins, the number of spares could be raised to 21.This issue of missing TX-plugins is discussed in more detail below.

The installation of the BOC cards in the ATLAS counting room USA15started in April 2007. 9 crates in 5 racks were equipped with 132 BOC cards.The distribution of the BOC cards on the different crates is determined by theATLAS pixel connectivity table [90]. One half of the BOC cards serves the Aside of the pixel detector, the other half serves the C side. This is true for allBOC card types (Layer-0, Layer-1, Layer-2, Disk-1,2,3). BOC cards for the Aand the C side are arranged alternatively in the crates instead of populating 1crate with A side BOC cards and the next with C side BOC cards. Table 4.6presents an overview of the installed BOC cards. Detailed up to date tables ofthe BOC card installation can be found in [91].

Table 4.6: Current status of BOC card installation in USA15.

Crate needed BOCs/RODs installed BOCs/RODs

Y25-05A2 upper 16 16Y25-05A2 lower 16 16Y26-05A2 upper 16 16Y26-05A2 lower 12 12Y27-05A2 upper 16 16Y27-05A2 lower 16 16Y28-05A2 upper 8 8Y28-05A2 lower 16 16Y29-05A2 upper 16 16

total 132 132

After installation of 1 TIM and 1 RCC (SBC) per crate, the installed BOCcards were powered up and showed a normal behavior of currents, voltages andtemperatures. The communication between the RODs and BOC cards workedas well. Due to clock problems, the BOC card 0x53 was replaced by the BOCcard 0x33, and the BOC card 0x49 was replaced by the BOC card 0x93. Dueto a broken S-Link connection, the BOC card 0x72 was replaced by the BOCcard 0x3A. More details about the commissioning of the BOC cards can be readin [92].

Since the installation in summer 2007 the BOC cards are operated. Occasion-ally, a laser channel on any TX-plugin dies (in figure 4.6 a photo of a TX-pluginis shown). This means that the emitted light power gets far below the necessarythreshold or is even not measurable anymore. This problem cannot be recovered,so the TX-plugin has to be changed. During roughly two years, more than 52TX-plugins had to be replaced. All information about failing TX-plugins such as


Table 4.7: Status of numbers of TX/RX-plugins for BOCs after their installation. The sumof “mounted and working”, “defect” and “spares” gives “existing”.

needed for existing mounted and defect sparesATLAS working

TX-Plugins 272 345 284 61 0RX-Plugins 316 366 348 4 14

detection date or serial number of the concerned TX-plugin were collected andare available in [93].On the BOC cards of the SCT detector some TX-plugins (that have 12 chan-nels instead of 8) failed with the same problem. Whereas, the failure rate ofthe RX-plugins is nearly zero for the pixel and the SCT detector. This meansthat there is a TX-plugin specific problem; perhaps the naked VCSEL array wasdamaged during production by electrostatic discharge (ESD). The problem isinvestigated amongst others in Taiwan, where the plugins have been produced,and in Wuppertal, where the plugins have been assembled. The status of the Tai-wanese TX/RX-plugins after the installation of the BOC cards is shown in table4.7. If the TX-plugin failure rate continues, not all channels of the pixel detectorcould be read out anymore, because spares are available in limited quantitiesonly. Therefore, it was decided to produce carefully new optimized TX-plugins.All existing plugins of the pixel and SCT detector have been replaced in sum-mer 2009, so that the optolinks are ready for the first beam collisions of the LHC.

4.3.2 Optical Cables

As mentioned in the overview of Section 4.2, optical fibers connect the ATLASpixel on-detector electronics (OBs) in the cavern UX15 to the off-detector elec-tronics (BOC cards) in the cavern USA15 [94]. More precisely, 80 m long glassfibers go from the BOC cards to the detector endplates at PP1. From there, fur-ther glass fibers with a length of 2.5 m go to the OBs at PP0. That is illustratedin figure 4.7. The Wuppertal group has taken a considerable responsibility forthe installation and commissioning of the optical cables.Due to the high levels of radiation inside the detector, when the LHC is in oper-ation, the fibers must be radiation hard. This can be achieved by using SteppedIndex Multi Mode (SIMM) fibers, which have a pure silica core. For the partoutside the detector, Gradient Index (GRIN) fibers are used, which are dopedand radiation tolerant. Corresponding to the connection layout in figure 4.7, the2.5 m long fibers between PP0 and PP1 are SIMM fibers. The fibers betweenPP1 and the BOC cards consist of two spliced fibers, one 8 m long SIMM fiberand one 72 m long GRIN fiber. Since one single fiber represents an unidirectional


BOCPP0

Off Detector

PP1

On Detector

Splicing Position

80m Optical Cable

8m

2.5m Optical Ribbon

Figure 4.7: The ATLAS pixel optical connection between OBs and BOC cards.

connection of one pixel module (MCC) to a ROD, the light signals can go eitherfrom the GRIN part into the SIMM part (“timing” fiber from the BOC card tothe OB) or from the SIMM part into the GRIN part (“data” fiber from the OBto the BOC card). For the “timing” case, the diameter of the GRIN part of thefiber is 62.5 µm. For the “data” case it is 50 µm. The diameter of the SIMM partof the fiber is always 50 µm. For more details look at [95].

Except for Layer-0, each of the 1744 modules requires 2 fibers, 1 to receive theclock/command signals and 1 to send the data signals. Layer-0 (for which thereare 286 modules) has an extra send fiber due to the higher occupancy. Thus,3774 fibers are needed in total. For handling reasons, the fibers are organizedas ribbons. On the BOC card side, 1 ribbon consists of 8 fibers, which areglued to each other in one row (Mechanical Transfer ferrule MT8 as terminatingconnector). On the on-detector side, 1 ribbon consists of 16 fibers in two rows(MT16 as terminating connector). This means that 2 ribbons on the BOC cardside form 1 ribbon on the PP1 side.

On the BOC card side, the male Sub Miniature C (SMC) connector, whichcontains the MT8 ferrule, of the ribbon is plugged into the SMC housing (TX-or RX-plugin) mounted on the BOC card. On the other side (at PP1) of the two80 m long MT8 ribbons, one male MT16 connector is connected to the femaleMT16 connector of the 2.5 m long MT16 ribbon. The OB has also a femaleMT16 connector.

Because of these MT8/MT16 and other modularities, 6, 7 or 8 fibers of oneribbon are used. Due to further modularities it even occurs 52 times that oneMT8 ribbon of two, both belonging to the same MT16 ribbon on the on-detectorside, is not used. Such a ribbon is called a “dark” ribbon. Furthermore, 32 spareMT8 ribbons were produced. In total, there are 336 MT16 or 672 MT8 ribbonsbetween the on- and off-detector side (this corresponds to 5376 fibers). Table 4.8presents an overview of the numbers of optical MT8 ribbons.

It would be too elaborate to lay 672 thin ribbons in the cavern. Furthermore,each time there would be a risk of damage. Therefore, 8 MT8 ribbons are arrangedtogether to form a cable. At the end of a cable, the last 3 m (1 m) of the 8 MT8(4 MT16) ribbons have a protection to be laid safely to their destinations. Insidethe cable no additional protection for each ribbon is necessary. So, there are 84


Table 4.8: Numbers of optical MT8 ribbons per destination and type (data/timing). Thesecorrespond to the numbers of needed BOC card plugins (RX/TX). Since 2 MT8 ribbons belongto 1 MT16 ribbon, the corresponding numbers of optical MT16 ribbons are half as much. Ofcourse, a ribbon can be laid only to one detector side (A or C), i.e. the corresponding numbersof MT8 ribbons for one detector side are half as much.

Destination Data (RX) Timing (TX) total

Layer-0 88 44 132Layer-1 76 76 152

Disk-1,2,3 48 48 96Layer-2 104 104 208

subtotal 316 272 588

dark 36 16 52spare 16 16 32total 368 304 672

cables in total (including spares). There are three types of cables: data, timingand hybrid, which contain 4 data and 4 timing MT8 ribbons. The numbers ofcables per type and detector side are patent from table 4.9.

The cables were laid according to the ATLAS pixel connectivity table [90].Thereafter, they were tested to check, if any of their fibers had been damaged.This was done by an Optical Time Domain Reflectometer (OTDR) which isshown in figure 4.8. Such a device can localize places in a glass fiber, where lightis more strongly reflected than on average. This can happen due to damage or aconnection of 2 fibers. The OTDR sends light into a fiber and measures the timet until a reflection comes back. So, the distance l to the reflecting place is givenby formula 4.1, where n is the refraction index of the glass fiber and c the speedof light.

l =c · t

2 · n (4.1)

The single mode (SM) OTDR, that was used, has a sensitivity of −110 dBm and

Table 4.9: Numbers of optical cables per type and detector side.

Data Cables Timing Cables Hybrid Cables total(yellow) (blue) (gray)

Side A 20 16 4 (+2 spares) 40 (+2)Side C 20 16 4 (+2 spares) 40 (+2)

total 40 32 8 (+4) 80 (+4)


Figure 4.8: Optical Time Domain Reflectometer (OTDR). [96]

operates at a wavelength of 1310 nm. The 1 point resolution is 5 mm, the 2 pointresolution is 10 cm. A computer (PC) steers the OTDR via an USB interface.The OTDR sends and receives light over a Stab and Twist (ST) connector.

The ribbons were tested by the OTDR from the off-detector side, where theyare more easily accessible. Since there are in one MT8 ribbon 8 fibers, whichshould be tested simultaneously in order to save time, a passive splitter was used.It splits the 1 channel output of the OTDR into 8 channels. To distinguish thesignals of the different channels, extension fibers are used so that each channelhas a different length. The passive splitter and the extension fibers have STconnectors, but the ribbon between on- and off-detector side has a SMC connectorat the BOC card side. So, a fanout (1 MT8 to 8 ST) was used. The MT connector(ferrule) of the fanout was connected via 2 steel guide pins to the SMC connectorof each ribbon. A sketch of the developed test setup is presented in figure 4.9.A measurement of a ribbon is shown in figure 4.10. On the x-axis the distancein meters is plotted and on the y-axis the relative backreflection signal in dBm.The OTDR is at 0 m. Every connector gives 1 peak. The marker A indicatesthe start of the first channel (fiber) of the ribbon, marker B indicates its end.In between, no peaks can be recognized. So, there should be no damages of theribbon. In figure 4.11 it is zoomed to the end of a ribbon (i.e. 80 m plus extensioncables/setup from the OTDR). All 8 peaks for the 8 channels can be clearly seen.This means that light is coming back from the end of the ribbon. Due to the2 point resolution of 10 cm there could be undetected damage inside the 10 cmfrom the PP1 end. This can be excluded only by connecting a further ribbon tothe far end or by testing also from the PP1 side.

After they have been laid, all 84 optical cables were tested in early 2007from the off-detector side (racks Y25-05A2 to Y29-05A2 in USA15). Only 4


PC

OTDR8 ST − 1 MT

7m

2 SMC − 1 MT16

80m

1−>8 channelpassive

splitter

1 ST − 1 ST

5m

USB

8 differentextensions

8 ST − 8 ST

Figure 4.9: Developed test setup for the 80 m long optical ribbons of the ATLAS pixel detectorbetween on- and off-detector side. The connector types and lengths of the ribbons/fibers arenoted. 2 MT8 ribbons (SMC connector), going to the same MT16 connector, are tested oneafter the other.

Figure 4.10: OTDR measurement of a ribbon.

Figure 4.11: OTDR measurement at ribbon end (PP1).


MT8 ribbons indicated problems (CR1-BAT12, CR4-YCD06, CR1-YAD16, CR2-YAD16). All of them do not show the expected reflection after 80 m (end ofribbon) for all channels. Furthermore, CR1-BAT12 shows an unexpected reflec-tion in all channels after 15 m from the BOC card side, CR4-YCD06 after 11 m,CR1-YAD16 and CR2-YAD16 after 9 m.Three of these 4 could be tested also from the on-detector side at PP1. Since theribbons have MT16 connectors at this end, a fanout ribbon with 2 MT8 connec-tors and 1 MT16 connector had to be inserted between the test setup and theribbon to be tested. The 3 investigated ribbons showed neither the expected re-flection after 80 m nor the unexpected reflections for all channels. So, these 4 MT8ribbons had to be replaced by spares. As 2 of the faulty ribbons have the sameMT16 end connector at PP1, only 3 of the 16 MT16 spare ribbons were needed(CR1-/CR2-BAT12 were replaced by CR1-/CR2-GCH04, CR1-/CR2-YAD16 byCR5-/CR6-GCH04, CR3-/CR4-YCD06 by CR5-/CR6-GCH05). Due to the de-scribed failures the ends of the optical cables were rearranged at the off-detectorside. 2 spare cables were laid to rack Y26-05A2 to replace the problematic rib-bons.

In summer 2007, the 3 m long ends of the ribbons at the off-detector side werelaid to their destinations (according to the results of the pixel connectivity testin the SR1-building [97] and the resulting connectivity table [98]) in the racksY25-05A2 to Y29-05A2 in USA15. Figure 4.12 shows a rack backside (BOC card

Figure 4.12: Back side of a rack with 2 crates in ATLAS counting room USA15. Left: afterinstallation of the optical cables but before installation of the optical ribbons. Right: afterinstallation of the optical ribbons.


side) before and after the installation of the ribbons. Every cable ends in one ofthe five racks, but each of the 8 ribbons of a cable can stay in the correspondingrack or go to the left or right neighboring rack. If a ribbon goes to a neighboringrack, it can go only to the upper (U28) crate, because the ribbon end is only 3 mlong. If a ribbon stays in that rack, where the corresponding cable ends, it cango to the upper (U28) or lower (U07) crate.Before the ribbons were connected to the BOC cards, all were tested again witha single mode OTDR as stated.Three ribbons showed problems in the OTDR tests:

1. Only fiber number 8 of ribbon CR1-BAT09 showed a reflection at 80 m. Allother 7 showed no reflection. So, this ribbon and ribbon CR2-BAT09 werereplaced by the spare ribbons CR3-/CR4-GCH04.

2. Fiber number 8 of ribbon CR8-YCD17 showed no reflection. No action wasnecessary as fibers number 1 and 8 are not used in this ribbon.

3. The two fibers number 7 and 8 of ribbon CR1-BCT03 showed no reflection.So, this ribbon and ribbon CR2-BCT03 were replaced by the ribbons CR1-/CR2-GAH05.

The reason for the malfunctioning of these 3 ribbons is unknown.A multi mode (MM) OTDR was used on some ribbons in order to confirm thesingle mode OTDR measurements. The results were as follows:

• “Low” reflections at 80 m in the single mode show “clear” reflections inmulti mode.

• Ribbons indicate no measurable attenuation (< 1 dB) over 80 m when all 8fibers are measured at the same time.

• When one single fiber of one ribbon is measured alone, a step in the at-tenuation (< 2 dB) can be observed at about 72 m (transition by splice ofSIMM and GRIN fibers). When 8 fibers are measured at the same time, thecorresponding 8 steps are not visible, because of the very low attenuationat the transition and of the overlap of the 8 light signals. A further reasonis that the position of the splicing is not excactly the same for all fibers.

• All other measurements done in single mode are confirmed by multi modemeasurements (especially missing reflections in fibers remain missing).

So, the connectivity table for the optical ribbons was updated and used for theconnection of the MT16 connectors at PP1 in early 2008. More details of theOTDR tests can be found at [99], all data of the described OTDR tests at [100].

During the connectivity tests [101] in March 2008, it turned out that theconnectivity table had a wrong mapping (swap) of the MT16 connectors at PP1 to


their corresponding 2 MT8 connectors at the BOC cards, so the 2 MT8 connectorswere swapped for all pairs. To get the correct mapping, some database entrieswere swapped. However, many MT8 ribbons had to be relaid in the racks ofUSA15. Fortunately, it seems (some OTDR checks were done again) that noribbon was damaged during this delicate work. More details about this recablingand the results of the connectivity tests are set forth in [102].

After the successful commissioning of the optoelectronics the pixel detectorwas commissioned and set into operation successfully. On 14 September 2008,the first cosmic muon was observed in the ATLAS pixel detector. Since then, thepixel detector is being calibrated — e.g. by the analysis of observed cosmic muonevents — and prepared for the data acquisition of LHC beam collision events.

Chapter 5

Event Reconstruction

In order to select and analyze collision events, the recorded raw data stored inRDO files, as introduced in Section 3.2, is used to obtain physics objects such astracks. This data processing step after the detection and recording of the collisionevents is called event reconstruction. The common objects needed for physicsmeasurements are tracks, vertices, jets, electrons, photons, muons and missingtransverse energy. Each of these objects and its reconstruction are summarizedin the following sections, more details can be found in [26, 34].

After the reconstruction of the physics objects, they are saved together withsome raw information in ESD files, which are mainly used for detector calibration.Physics analyses such as top quark precision measurements are more suitablyperformed with AOD or DPD files, which are derived from ESDs and small insize, because they contain almost no raw information. More information aboutthe general reconstruction step in ATLAS can be found in [103]. In Section 7.1,the corresponding processing steps are explained for Monte Carlo simulation.

5.1 Tracking

The muon spectrometer allows a track reconstruction over the range |η| < 2.7 formuons with momenta between 3 GeV and 3 TeV. The inner detector contributesalso to the measurement of muons. Moreover, vertices and tracks are recon-structed there, which are important for the jet tagging as we will see in Section 6.1.Typically, charged particle tracks with transverse momentum pT > 0.5 GeV and|η| < 2.5 are reconstructed in the inner detector.

The inner detector track reconstruction software [104] is responsible to recog-nize tracks in a large number of measured space points. Tracks represent chargedparticles traversing the detector and depositing energy (“hits”). The software usesgenerally for track fitting global χ2 and Kalman filter techniques. Furthermore,dynamic noise adjustment [105], Gaussian sum filters [106] and deterministic an-nealing filters [107] may be applied, especially for electron tracks. Other tools

52 Chapter 5. Event Reconstruction

such as those for track extrapolation are provided by the inner detector trackreconstruction software as well.

Before tracks can be found or fitted, the raw data from the pixel and SCT de-tectors is converted into clusters, and the TRT raw timing information are turnedinto calibrated drift circles. The default strategy to find tracks works outwards:Space points in the three pixel layers and the first SCT layer are combined tomake track seeds, which are extended throughout the SCT to form track candi-dates. These candidates are fitted, outlier clusters are removed, ambiguities inthe cluster-to-track association are resolved, and fake tracks are rejected. Then,the selected tracks are extended into the TRT to associate drift circle information.Furthermore, left-right ambiguities are resolved. Finally, the extended tracks arerefitted with the full information of the pixel, SCT and TRT detectors. The qual-ity of the refitted tracks is compared to the silicon-only track candidates. Hitson track extensions resulting in bad fits are labelled as outliers.

Muon tracks can be reconstructed by the muon spectrometer data only (stand-alone fit), by the combination of the two tracks from the muon spectrometer andthe inner detector (combined fit) or as a track from the inner detector to an innermuon station (segment fit). Similar to the strategy for inner detector tracks, theraw data of the muon spectrometer is first preprocessed to form drift circles orclusters. Then, patterns are found, segments are made and combined, and finallytracks are fitted.

However, the tracking may not reconstruct the tracks of all charged parti-cles that traverse the detector. Furthermore, some tracks may be reconstrucedthough no corresponding particle travelled there. Such tracks, which pass somequality cuts but are not matched to any particle, are called fake tracks. Theratio of such tracks compared to all reconstructed tracks is the fake rate. Thisrate can be estimated by MC simulation: Generated MC particles traverse thesimulated detector, and the corresponding detector response is simulated. Thereconstruction software runs over this data in the same way as it is done for realdata. Then, the reconstructed tracks are matched to MC particles, if at least80% of their detector hits were created by a MC particle. The remaining tracksthat have no match are fake tracks.The standard quality cuts of the tracking require at least seven precision hits inthe pixel or SCT detector. The fraction of MC particles that are matched toreconstructed tracks passing these quality cuts is called the tracking efficiency.Stricter selection cuts are used for the tracks in the b-tagging and require at leasttwo hits in the pixels and one in the vertexing layer. For example, the efficiencyis decreased and the fake rate is increased by cavern background and pile-up.

The track resolution strongly depends on the alignment of the detectors andthe determination of the magnetic field. The first affects more the inner detector,whereas the second is more important for the muon spectrometer.The magnetic field in the muon spectrometer is continuously monitored by 1730Hall cards [38, 108], which make a 3D measurement. The absolute Hall-card

5.1 Tracking 53

accuracy on | ~B| is 0.2 mT up to | ~B| = 1.4 T and 1 mT up to 2.5 T; the angularaccuracy achieved on the measured field direction is 2 mrad. The field of thesolenoid was mapped [109] before the inner detector was installed but with allcalorimeters in their final positions. In order to monitor the ID field strengththroughout the lifetime of ATLAS, the inner detector is equipped with four NMRprobes. These are fixed to the wall of the inner warm vessel near z ≈ 0 and equallyspaced in azimuth. They measure the field strength with an accuracy of about0.01 mT.

The alignment procedure of the inner detector determines the positions inspace of the silicon modules (1744 pixel and 4088 SCT modules) as well as of thestraws in the 136 TRT modules every 24 hours. Therefore, a system with almost3.6 · 104 degrees of freedom (1744 + 4088 + 136 = 5968 modules times 3 + 3 = 6shifts/rotations) has to be solved. With one million good tracks in this period theprecision on the silicon module positions may reach 10 µm. Track based alignmenttechniques use the minimization of hit residuals from high momentum tracks.Since multiple scattering decreases with increasing momentum, high momentumtracks have lower distortions and are well suited for the purpose of alignment. Inorder to determine all global distortions, tracks with different topologies are used,for example: tracks from the interaction point, cosmic ray tracks, tracks frombeam halo, and tracks passing through the overlap regions of adjacent modules.

The absolute positions of the muon chambers are also determined by trackbased techniques. Whereas, the relative positions of the precision chambers arecontinuously monitored by optical in plane alignment sensors. Thereby, an accu-racy of ≤ 30 µm can be achieved for the chamber locations [26].

Since the magnetic field is not perfectly homogenous and the track particleslose energy by traversing the detector material, a track is not an exact helix.Normally, the track parametrization is chosen in a frame, where the z-axis isparallel to the magnetic field. Figure 5.1 illustrates the five track parametersused in ATLAS:

• the charge over momentum magnitude q/p

• the transverse impact parameter d0, which is the distance of the closestapproach of the track projected into the x−y plane relative to the nominalinteraction point (0, 0, 0). During track reconstruction primary verticescannot be chosen as reference points, because they are found and fittedafterwards.

• the longitudinal impact parameter z0, which is the z value of the point ofclosest approach determined as above

• the azimuthal angle φ0 of the momentum at the point of closest approachdetermined as above. φ0 is measured in the range [−π, π).

• the polar angle θ in the range [0, π]


Figure 5.1: The five track parameters, split into three transverse ones (x − y plane) and twolongitudinal ones (r − z view) for a helix track in a cylindrical detector geometry. [110]

For the consideration of their resolutions, three of the track parameters are modi-fied as q/pT , cot(θ) = pz

pT, z0 · sin(θ) (z0 projected onto the plane transverse to the

track direction), because then they are more useful for many applications such aselectron identification or particularly jet tagging. However, the multiple scatter-ing of a charged particle makes the resolution of the track parameters worse thanthe intrinsic detector resolution. Since the multiple scattering decreases withlarger momentum, the resolution of the track parameter X can be estimated as:

σX(pT ) = σX(∞) (1 ⊕ pX/pT ) (5.1)

The resolution σX(∞) is the asymptotic resolution expected at infinite momen-tum. The parameter pX is a constant for the parameter X under considerationand represents the value of pT , for which the intrinsic and multiple scatteringterms are equal. Table 5.1 provides the expected resolutions of the five trackparameters according to the mentioned detector layout and accuracies of thealignment as well as of the magnetic field. We will see in Chapter 6 that an ac-curate track reconstruction and high resolutions of their impact parameters arevery important for the selection, identification and analysis of top quark events.

5.2 Vertexing 55

track parameter0.25 < |η| < 0.50 1.50 < |η| < 1.75

σX(∞) pX [GeV] σX(∞) pX [GeV]inverse transverse

0.34 TeV−1 44 0.41 TeV−1 80momentum q/pT

azimuthal angle70 µrad 39 92 µrad 49

φpolar angle

0.7 · 10−3 5.0 1.2 · 10−3 10cot(θ)

transverse impact10 µm 14 12 µm 20

parameter d0

longitudinal impact91 µm 2.3 71 µm 3.7

parameter z0 · sin(θ)

Table 5.1: Expected track parameter resolutions (RMS) at infinite transverse momentum,σX(∞), and transverse momentum, pX , at which the multiple scattering contribution equalsthat from the detector resolution. The values are shown for two η regions, one in the barrel ofthe inner detector, where the amount of material is close to its minimum, and one in the endcap, where it is close to its maximum. [34]

5.2 Vertexing

When the reconstruction of tracks is finished, vertices can be found and fitted,which represent the positions, where particles decay. Especially the reconstruc-tion of primary vertices, where interactions between beam particles occured, andconversions of photons into positron-electron pairs in the detector are importantfor many physics analyses.

There are several algorithms to find and fit primary vertices [34,111,112]. Allprimary vertex finders use tracks, which originate from the beam crossing area.The expected resolution of the primary vertex in top-antitop events is 18 µm in thex−y plane and 41 µm in the z-direction. Due to a beam constraint, the resolutionin the transverse plane can be improved to 11 µm [34]. The situation whencollision events have more than one primary vertex is called pile-up, as alreadyintroduced in Section 2.3. Especially in such a case a good tracking performanceis important in order to associate the tracks to the primary vertex they originatefrom. A correct association reduces backgrounds of physics analyses and is vitalfor the b-tagging of jets, as we will see in Chapter 6.

Secondary vertices represent particles decaying at a point displaced from anyprimary vertex. Especially when a particle containing a b-quark decays, thereconstruction of the corresponding vertex may be important to identify a jetas a b-jet. The resolution of the radial position of secondary vertices stronglydepends on where the particle decays: if the decay is in front or behind of adetector or detector layer may cause a factor of up to 10.


5.3 Electrons and Photons

Electrons and positrons are both referred as electrons due to their undistinguish-able behavior in the calorimeter. The inner detector may determine the sign ofthe electric charge of an electron measured in the calorimeter, if an electron trackcan be associated to it.

For the reconstruction of tt events with one top quark decaying leptonicallyin the electron channel, it is crucial to find the electron in the large amountof jets: at the LHC the electron-to-jet ratio is expected to be about 10−5 atpT = 40 GeV [34]. Moreover, the accurate reconstruction of electrons and theirdiscrimination from photons are important for precision measurements.The reconstruction and identification of electrons/photons starts with a seedcluster from the electromagnetic calorimeter. Electron and photon candidatesare separated reasonably cleanly by requiring the electrons to have an associatedtrack but no associated conversion into an electron/positron pair. Tracks andconversions are both reconstructed in the inner detector. Similarly, photons haveno matched track but may have a matched reconstructed conversion. Showershape variables, for example lateral and longitudinal shower profiles, help toseparate electrons from jets. Finally, calorimeter energy isolation or observablesfrom the inner detector such as TRT high threshold hits are used to identifyelectrons and photons.

These observables from the inner detector and from the calorimeters also helpto reject jets. The corresponding rate is between 500/1300 for low energeticelectrons with loose/medium cuts (shower-shape or track cuts) and 105 for highenergetic electrons with tight cuts (high ratio between high-threshold and low-threshold hits in the TRT detector or calorimeter energy isolation). However, theefficiency strongly depends on the physics process, for example the efficiency of“tight isolated” high-energetic electrons may be 64% for Z → ee but only 16%for electrons, which are near heavy flavor jets [34].

For photons the jet rejection rate is not as high, because jets can fake photons,especially when they contain single or multiple neutral hadrons. Nevertheless, aquark jet rejection rate of about 1500 without and 3000 with track isolation isachievable for high energetic photons. For gluon jets these numbers are by afactor of 10 larger, because they have a broader lateral extent due to the softerfragmentation [34].

Thus, the inner detector contributes a lot to the reconstruction and identifi-cation of electrons/photons.

5.4 Missing Transverse Energy

In order to measure the top quark mass from tt events, in which one top quarkdecays leptonically, a good measurement of the missing transverse energy /ET in

5.5 Jets 57

terms of linearity and accuracy is important. /ET is determined by the muonspectrometer and the calorimeters using the same global calibration of calorime-ter cells as for jets. The uncalibrated /ET is the energy that sums up with thetransverse energy of all calorimeter cells and the pT of the muons, measured by themuon chambers, to zero transverse energy. To avoid fake muons, only muons witha track match in the inner detector are selected. Since a lot of energy is lost in thecryostat between the barrel LAr electromagnetic and tile calorimeter, a correctionhas to be applied, which represents about 5% per jet with pT > 500 GeV [34].A refined calibration of /ET is obtained, if each reconstructed high pT object isassociated to its globally calibrated cells.

Over a wide range of the total transverse energy deposited in the calorime-ters, the resolution σ of the missing transverse energy follows an approximatestochastic behavior: A simple fit to the function σ = a ·

√

∑

ET yields valuesof about 0.55 ± 0.02 for the parameter a, when

∑

ET is in the range between20 and 2000 GeV. Above 2 TeV, the constant term in the jet energy resolutiondominates, as we will see in Section 5.5. Monte Carlo simulations provide anestimation of the azimuthal accuracy, which is defined as the angular differencebetween the directions of reconstructed and true missing transverse energy. Fora missing transverse energy above 100 GeV a resolution of about 100 mrad is ex-pected.However, several effects may change the direction and amount of the missingtransverse energy: beam gas scattering and other machine backgrounds, dis-placed interaction vertices, hot/dead/noisy cells (or regions) in the calorimeters,and mismeasurements in the detector itself, for example due to high pT muons es-caping outside the acceptance of the detector and due to large losses of depositedenergy in cracks or inactive material. Normally, the transverse missing energyis increased by such effects. Therefore, the fake /ET is defined as the differencebetween reconstructed and true missing transverse energy. The ratio betweenfake and true /ET can be reduced by applying cuts. For QCD dijet events a re-duction of a factor 10 over the whole energy range can be gained, if the distancein azimuth between the reconstructed /ET vector and the direction of any high pT

jet is required to be larger than 17◦. [34]

5.5 Jets

At the LHC, jets will be present in almost every event, whether they originatefrom QCD background processes or represent signal, for example in top quarkpair events. As introduced in Section 3.2.2, jets are detected by the calorimeters.However, there is a long chain of processing steps from calorimeter signals to jets,which are finally calibrated to parton level, as illustrated in figure 5.2.

Jets are formed from signals of all calorimeters. For a fast jet finding, cellsignals are combined into clusters. There are two possibilities for such a combi-


nation: tower clusters and topological clusters. Towers are formed by collectingcells into bins of a regular ∆η ·∆φ = 0.1 · 0.1 grid and summing up their signals.If a cell overlaps partly with the tower bin, a fraction of its signal is added, whichcorresponds to the overlap area fraction.Topological cell clusters represent an attempt to reconstruct three-dimensionalenergy depositions in the calorimeters. Starting from seed cells, which have ahigh absolute signal, close cells are collected, if they have absolute signals abovecertain thresholds.

There are a lot of jet-clustering algorithms, which reconstruct jets from thetowers or topological clusters. Two of them are very common in ATLAS: a seededfixed-cone algorithm and a successive recombination algorithm (kT ). Both canproduce narrow (∆R = 0.4) or wide (∆Rcone = 0.7, ∆RkT

= 0.6) jets. The conealgorithm is the default for the high-level trigger, because it is very fast. Asparameters it uses the transverse energy threshold for a seed and the cone size∆R. A split-and-merge step follows the actual cone building, with an overlapfraction threshold of 50%. The successive recombination algorithm is preferredfor offline physics analyses. However, the choice of the jet algorithm depends alsoon the physics channel under consideration.

After the jets are found, they have to be calibrated. First, the cell signals areweighted to obtain the hadronic calibration. The standard approach is similar tothe one developed for the H1 calorimeter [113]. The weighting functions dependon the calorimeter signal types (towers or topological clusters), on the jet algo-rithm (for example cone or kT ) and its specific configuration, and on the physicscalibration samples. Therefore, jet energy scale corrections have to be applied toget physics jets, which are calibrated to particle level.Finally, physics jets are refined to interaction/parton level by in-situ measure-ments, which include for example pile-up. Good choices for this task are finalstates such as γ+jet(s), Z+jet(s) or hadronically decaying W bosons (W → qq),because they have a well measured electromagnetic object, which balances one ormore jets in transverse momentum or because the mass of the W boson constrainsthe energy scale of the two quark jets.

In order to analyze the performance of the jet reconstruction, efficiency andfake rate are determined in a similar way as described in Section 5.1. Gener-ated MC particles are clustered, usually by cone algorithms, into “truth” jets.Therefore, the jet reconstruction efficiency is defined as the ratio between thenumber of matches of “truth” jets with reconstructed jets and the total numberof “truth” jets. Whereas, the purity of the jet reconstruction can be expressedas the ratio between the number of matches of “truth” jets with reconstructedjets and the total number of reconstructed jets. The fake rate is then one minuspurity. The corresponding values depend on the chosen matching radius, whichis typically ∆Rm = 0.2. Moreover, the two calorimeter signal definitions (towersor topological clusters) as well as the various jet algorithms may lead to differentefficiencies or purities. Finally, pile-up and electronic noise may decrease the effi-

5.5 Jets 59

ciency. However, efficiency and purity are almost 100% for jets with a transverseenergy above 50 GeV.

In order to analyze efficiencies and purities of jet reconstruction and jet tag-ging in more detail, a jet label is defined, which uses MC information. This labelindicates from which type of parton the jet originates. However, this labellingprocedure is ambiguous. Usually, a jet is labelled as a b-jet, if a b-quark withtransverse momentum pT > 5 GeV is found in a cone of size ∆R = 0.3 aroundthe jet direction. If there is no b-quark found but a c-quark or a τ -lepton, whichfulfills the same conditions, the jet is labelled accordingly. Otherwise, the jetis labelled as a light jet. Especially in Chapter 6, these labels are needed forperformance studies.

After global calibration the fractional jet energy resolution, whose perfor-mance is quite important for the reconstruction of tt events, may be fitted byfollowing function:

σ

E=

√

a2

E+

b2

E2+ c2 (5.2)

For cone jets made from towers with ∆R = 0.7 and 0.2 < |η| < 0.4, the stochas-tic term [a/

√E] is ≈ 60%

√GeV, while the high-energy limit of the resolution,

expressed by the constant term c, is ≈ 3%. The parameter b represents the noise,which increases from 0.5 GeV to 1.5 GeV when going from the barrel to the endcap.

Another important observable of the jet reconstruction is the jet axis. AMC study shows that the choice of the calorimeter signal (towers or topologicalclusters) does not significantly affect the reconstructed jet direction. Though, thereconstruction of the jet axis is degraded at transverse energies below 100 GeV.However, the precise measurement of the jet direction is very important for b-tagging, as we will see in the next chapter.


Figure 5.2: Jet reconstruction flow for calorimeter jets from towers or clusters. [34]

Chapter 6

Tuning of Track ImpactParameters for b-Tagging

Track impact parameters are important for b-tagging as we will see in Section 6.1.In order to understand the detector and to have reliable measurements of impactparameters, it is necessary to perform simulations. If these are in agreement withdata, maximal systematic uncertainties can be estimated. However, track impactparameter distributions will quite likely not agree between first data and MonteCarlo simulation because of a misalignment of the detector, detector inefficiencies,etc. Therefore, the MC distributions are tuned to account for such effects. Afterthe tuning, the deviations between simulation and data distributions providean estimate of the systematic uncertainties of track impact parameters. Thischapter overviews the b-tagging, presents the development and tests of an impactparameter tuning approach, and points at eventual alternatives.

6.1 b-Tagging

For many physics analyses like Higgs boson searches or top quark measurements itis important to identify jets stemming from the fragmentation and hadronizationof b-quarks. Hence, there are jet tagging algorithms, which calculate for each jeta probability that the jet originates from a certain quark flavor.

b-tagging algorithms use the fact that b-jets contain a B-hadron, that hasa relatively long lifetime of the order of 1.5 ps (c · τ ≈ 450 µm). A B-hadronin a jet with pT = 50 GeV will therefore have a significant flight path length〈l〉 = β · γ · c · τ [114], typically 3 mm in the plane transverse to the beambefore decaying. The tracks from the B-hadron decay products can be used toreconstruct the decay vertex, a so called secondary vertex (SV). If such a vertex isnear or inside a jet, this jet may be tagged as a b-jet. A jet can be also identified byusing the measured impact parameters (IP) of its tracks. Whereas in the tracking,the track parameters are usually defined relative to the nominal interaction point

62 Chapter 6. Tuning of Track Impact Parameters for b-Tagging

Figure 6.1: Example of a positive jet tag: The jet may be tagged due to the reconstructedsecondary vertex, which lies near the jet axis, or the large impact parameters of its tracks. [115]

(0, 0, 0), the jet tagging uses definitions relative to the associated reconstructedprimary vertex: The transverse impact parameter, d0, is the distance of closestapproach of the track to the primary vertex (PV) point in the r − φ projection.The longitudinal impact parameter, z0, is the z coordinate of the track at thepoint of closest approach in r−φ. [114] As illustrated in figure 6.1, tracks from a B-hadron decay tend to have large impact parameters which can be distinguishedfrom small impact parameters of tracks originating from the primary vertex.Therfore a jet, to which large impact parameter tracks are associated, can beidentified as a b-jet. As a consequence, b-tagging relies highly on tracking andvertexing.

For further discrimination of tracks from B-hadron decays from tracks orig-inating from the primary vertex the impact parameter is signed. As illustratedin figure 6.2, the sign is positive, if relative to the primary vertex the point ofclosest approach lies on the side of the jet; otherwise the sign is negative. Moreprecisely, the sign is defined using the jet axis ~Pj as measured by the calorimeters,

the direction ~Pt and the position ~Xt of the track at the point of closest approachto the primary vertex and the position ~Xpv of the primary vertex [114]:

sign(d0) = sign[

(~Pj × ~Pt) ·(

~Pt × ( ~Xpv − ~Xt))]

(6.1)

Similarly, the sign of the longitudinal impact parameter z0 is given by the sign of(ηj − ηt) · z0t, where ηj is the pseudorapidity of the jet, ηt is the pseudorapidityof the track and z0t is the impact parameter relative to the primary vertex.

6.1 b-Tagging 63

��

��

��

��

Track 2

IP>0 Track 1

PV

Reconstructed Jet Axis

IP<0

Figure 6.2: Visualization of the definition of the signed track impact parameter.

Due to the limited experimental resolution the sign of the impact parametersof tracks originating from the primary vertex, whatever jet they are associatedto, is random in an ideal experiment. However, particles can be deflected bymaterial interactions. If this deflection is not corrected for, the reconstructedtrack may have a positive, larger impact parameter than the particle would havewithout any material interaction. Similarly, impact parameters may be shiftedtoward higher values, if some decay vertices are not resolved. Hence, the impactparameter distribution of tracks associated to the primary vertex has larger tailson the positive side.

The signed impact parameter distribution of tracks is even more asymmetricbecause of long lived particles (e.g., K0, Λ0, D0, B0, etc.), that decay displacedfrom the primary vertex and lead to much more entries on the positive side of thedistribution. Although B-hadrons have a shorter lifetime than lighter hadrons,the positive side of the signed impact parameter distribution is more raised forb-jets than for light jets, as illustrated in figure 6.3. This is due to the proportionof tracks originating from long lived particles. Though the negative side of theimpact parameter distribution is mainly determined by the intrinsic detectorresolution, negative tails arise, if the reconstructed jet axis deviates from theprimary hadron flight axis. This is illustrated in figure 6.4 and happens especiallyin b-jets, when the distance between B- and C-hadron decay vertices is significantcompared to the vertex resolution in flight direction. The larger the deviationis, the more sizeable are the negative tails. If a track from the B-hadron decayvertex lies between both axes (straight track assumed), it has a negative impactparameter.

Due to multiple scattering, the track impact parameter distribution is a func-tion of the transverse momentum pT . There is also a dependence on the pseudora-pidity η of the track, because of the geometry of the detector (different resolutionsand distances between measurement points) and the distribution of the material.These dependencies are illustrated in figures 6.5 and 6.6. Similarly, the errorof the impact parameter can vary significantly and depends on pT , η, clustertopology and hit pattern. In order to give more weight to precisely measured


Signed Transverse Impact Parameter [mm]

-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110light jet

b-jet

Figure 6.3: Signed transverse track impact parameter distribution for b- (red) and light (blue)jets in tt events.

tracks, the impact parameter IP is divided by its error σIP. Consequently, abetter discrimination is reached, whether a track stems from the primary vertexor not. This discriminating variable d0/σd0

or z0/σz0is called the signed impact

parameter significance.

Track impact parameter significances are used by several b-tagging algorithms,for example IP1D (z0/σz0

), IP2D (d0/σd0) and IP3D (combination of IP1D and

IP2D) [114]. The first two algorithms use a likelihood ratio method: The sig-nificance Si of a track i, e.g. d0/σd0

, is compared to predefined smoothed andnormalized distributions (so called reference histograms) for both the b- and lightjet hypotheses, b(Si) and u(Si). The ratio of the probabilities b(Si)/u(Si) definesthe track weight, which can be combined into a jet weight Wjet as the sum of

��

��

��

��

��

��

PVIP<0 IP>0

ReconstructedJet Axis 1

Track

Primary Hadron/Reconstructed Jet Axis 2

Figure 6.4: The sign of the track impact parameter depends on the reconstruction of the jetaxis. Jet axis 2 (green) may be the flight direction of the B-hadron, and the sign would bepositive, but jet axis 1 (purple) is reconstructed and this leads to a negative sign.

6.1 b-Tagging 65


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 < 2 GeV

T1 GeV < p

< 3 GeVT

2 GeV < p

< 4 GeVT

3 GeV < p

T4 GeV < p

Figure 6.5: Dependence of the track impact parameter on the transverse momentum of thetrack. The tracks are associated to any jet of tt events.

the logarithms of the Nt individual track weights Wi:

WJet =Nt∑

i=1

ln(Wi) =Nt∑

i=1

ln

(

b(Si)

u(Si)

)

(6.2)

To select b-jets, a cut value on WJet must be chosen and corresponds to a givenb-tagging efficiency, which is defined as follows: Out of all jets labelled as b-jets(according to the jet labelling explained in Section 5.5) how many of them aretagged as b-jets? The performance of a b-tagging algorithm is then given by itsefficiency and its rejection rate, that is the inverse of the mistagging rate. Similarto the efficiency, the mistagging rate answers: out of all jets labelled not as b-jets, how many of them are tagged as b-jets? However, the efficiencies and thecorresponding rejection rates of many b-tagging algorithms depend for exampleon the jet transverse momentum and pseudorapidity.

Some algorithms such as IP2D require an a priori knowledge of the prop-erties of both b- and light jets, namely the reference histograms of the trackimpact parameter significances. The b-tagging algorithms were implemented andcalibrated by using MC simulation data samples. But these samples may havedifferent distributions of track parameters (d0, z0, φ, θ, q/p and their errors orthe corresponding error matrix V) than real data samples. Some possible reasonsare: misalignment of the detector, detector inefficiencies and detector materialdescription; different heavy flavor content for different generators of the simula-tion. A different impact parameter distribution would lead to a change of the



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110

-2.5 < eta < -1.9

-1.9 < eta < 0

0 < eta < +1.9

+1.9 < eta < +2.5

Figure 6.6: Dependence of the track impact parameter on the pseudorapidity of the track.The tracks are associated to any jet of tt events.

tagging efficiency and therefore of the signal amplitude of the physics process inconcern. To get closer agreement with data, the MC generator or the detectorsimulation may be tuned. Such a procedure can be quite complicated and elabo-rate. Instead, the track impact parameters of the MC samples may be adjusted,a procedure which can be done more rapidly. A possible approach of such amethod is presented in the next section.

6.2 Adjustment Strategy

To get closer agreement with data, the MC track impact parameter distributionmay be forced to the shape of the real data (RD) distribution, as illustrated infigure 6.7. The reconstructed MC impact parameter dMC

0 (likewise zMC0 ) is the

sum of the true impact parameter dMC,true0 and its correction ∆dMC

0 :

dMC0 = dMC,true

0 + ∆dMC0 (6.3)

The correction ∆dMC0 depends mainly on the detector resolution, and the true

impact parameter is zero (dMC,true0 = 0 mm) for tracks (simulated particles) from

the primary vertex. When the distributions of such tracks do not coincide for MCand RD, the MC correction (∆dMC

0 ) has to be adjusted for all MC tracks. Afterthe MC correction is set to the corresponding RD resolution (∆dMC

0 := ∆dRD0 ), the

distribution of the adjusted MC impact parameter (dMC,AD0 ) should be mapped

6.2 Adjustment Strategy 67

Figure 6.7: Sketch of the mapping approach. A MC track has an impact parameter of−0.36 mm. This corresponds to a position in the distribution, where the integral is 30% (wholedistribution normalized to 1). The position in the real data distribution (also normalized to 1),where the integral is also 30%, is −0.40 mm. So, the MC track impact parameter is forced tothe value −0.40 mm.

on the RD distribution (hence the name mapping approach) and they shouldagree well.

Using the RD impact parameter distribution, the resolution ∆dRD0 is obtained

as follows: The negative side of the RD impact parameter distribution (dRD0 ) pro-

vides a good estimation of the resolution (∆dRD0 ), because it is dominated by the

intrinsic detector resolution, as indicated in the previous section. When a sym-metric resolution around zero is assumed, the full resolution distribution (∆dRD

0 )is obtained by just mirroring the negative side to the positive. For a certain MCtrack the corresponding RD resolution (∆dRD

0 ) is defined, when the integral overthe normalized MC distribution (obtained by the same mirror technique as forRD) up to the MC correction value (∆dMC

0 ) is equal to the integral over the RDdistribution (for both distributions the lower/upper integration limit is ±1 mm,which is the default impact parameter cut for b−tagging in ATLAS [114]):

∆dMC0

∫

−1

1

N

d n

d dMC0

d dMC0 =

∆dRD0

∫

−1

1

N

d n

d dRD0

d dRD0 (6.4)


Then, the adjusted MC impact parameter for tracks from the primary vertex is:

dMC,AD0 = dMC,true

0 + ∆dMC,AD0 = 0 + ∆dRD

0 (6.5)

Since the track impact parameter distribution depends on observables like thetransverse momentum pT (due to multiple scattering) or the pseudorapidity η(due to detector geometry) of the track (as shown in figures 6.6 and 6.5), theadjustment has to be made in bins of these observables.

For tracks, which do not originate from the primary vertex, the adjustmentstrategy is the same. Since it is assumed, that the resolution does not stronglydepend on the track type (from primary vertex or not), ∆dRD

0 can be retrievedin the same way as above. The true impact parameter dMC,true

0 is not zero, butthe equations 6.3, 6.4, 6.5 still hold.However, the following two approximations for both types of tracks are reasonableas the results below indicate: The true and the reconstructed primary vertex areexactly coinciding. The true and the reconstructed impact parameter, which isactually a vectorial quantity, are parallel.

Most b-tagging algorithms do not use the track impact parameter itself butits significance. When the reconstructed MC impact parameter is adjusted as ex-plained above, the significance will change accordingly. For b-tagging algorithmsrelying only on track impact parameters the efficiencies and rejection rates mayagree between MC and data, after the adjustment was done.However, it is still possible, that the impact parameter significance distributionof MC does not coincide with that of data. A reason may be an unaccurateestimation of the individual track parameter errors. Then, it may be applicativeto scale the error of the impact parameter as well. A first approach is presentedin [116].

The implementation of the track impact parameter mapping approach withinthe Athena framework of ATLAS is described in Appendix A.1.

6.3 Performance and Tests

The performance tests of the presented approach use the default tracking [104],jet definition and reconstruction [114] of ATLAS. During these studies LHC andATLAS had not collected any collision data. But the procedure of tuning trackimpact parameters needs to be tested before collision data is available. Insteadof using MC and real data, one can use two different MC data samples. Asample with a perfectly aligned detector geometry may act as MC data, whereasa MC data sample with a misaligned detector geometry can act as real data ‘RD’(pseudo data). In figure 6.8 signed track impact parameter distributions of suchsamples and their difference are shown. As expected, the distribution of the MCdataset, which has a misaligned detector geometry, is broader (corresponding to

6.3 Performance and Tests 69


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.005

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

Figure 6.8: At the top: signed transverse track impact parameters for MC and pseudo realdata (“Real Data”); At the bottom: difference of MC and pseudo data.


a worse resolution) than the one having a perfect geometry: at small impactparameters, the difference between MC and pseudo data is positive.

A first test of the presented method is illustrated in figure 6.9: a Gaussian dis-tribution (black) is mapped (red) to another Gaussian distribution (blue) havinga different width. As expected, the blue and the red distribution are coincidingvery well.

Figure 6.9: First test of the method: a gaussian distribution (black) is mapped (red) toanother gaussian distribution (blue).

Since b-tagging is very important for the selection and analysis of top quarkevents, especially precision measurements of top quark properties, the mappingapproach was applied using two top quark pair datasets. Their generated eventsare the same, but they differ in the alignment used for the reconstruction. Theone serving as MC [117] has 6 · 105 events, the other serving as pseudo RD [118]has 5 ·104 events, which is a factor 12 less. In order to provide similar conditions,the MC and ‘RD’ distributions (“mapping histograms”), which are used to de-termine ∆dRD

0 for each MC track, have the same statistics of 5 · 104 events. Thebin size of the histograms is chosen as 20 µm, so that it is a bit larger than theexpected resolution of the transverse impact parameter, see table 5.1. Since thesigned impact parameters depend on the transverse momentum pT and pseudora-pidity η of the track as explained in Section 6.1, several mapping histograms werecreated for different pT and η ranges: 4 bins are used for the pT (in units of GeV):1 − 2, 2 − 3, 3 − 4 and > 4. In η the following 5 bins were chosen: (−2.5;−1.9),(−1.9;−0.1), (−0.1; +0.1), (+0.1; +1.9), (+1.9; +2.5). As a result, there are 20mapping histograms, both for the MC and the pseudo RD dataset. Moreover,

6.3 Performance and Tests 71

these distributions were created for the signed transverse track impact param-eter d0 as well as the longitudinal z0. Apart from transverse momentum andpseudorapidity the impact parameter distribution depends also on other trackproperties. For example, tracks that share some of their hits with other tracksrepresent a set yielding a worse resolution. Such effects are not considered in thisfirst approach.

Figure 6.10 shows that a statistics of about 5 · 104 tracks is sufficient fora successful mapping in the range of |d0| < 0.4 mm. When the two differencedistributions before and after adjustment (pseudo RD − original MC versus ad-justed MC − pseudo RD) are compared, it can be seen that the agreement inthis range is up to a factor of 4 (| − 0.008/4| = 0.002) better after the map-ping procedure. However, the adjustment overshoots in this region, becausethe sign of the difference inverts. Up to now, the reason for this behavior isunknown. For large impact parameters there is nearly no improvement be-cause of the small number of tracks, which are available for the mapping his-tograms. The statistics in the range |d0| > 0.4 mm is two orders of magnitudelower than for small impact parameters. Furthermore, the statistics is limitedby the pseudo RD sample, which has 12 times fewer events than the MC one.

The results are confirmed by further plots provided in Section A.2. In particu-lar, the argument that a failing mapping is caused by low statistics is supported:At high pseudorapidities (1.9 < |η| < 2.5) the mapping leads to smaller im-provements than at low pseudorapidities (0.1 < |η| < 1.9), where the number ofavailable tracks is about 6 times larger.

Section A.2 provides all 40 histograms, which present the results of the map-ping approach done for the chosen datasets: for the transverse and longitudinalsigned impact parameter, for the 5 bins in the pseudorapidity and the 4 bins inthe transverse momentum of the track.

Since the mapping approach leads to a better agreement between MC andpseudo RD by up to a factor of 4, it is a good candidate to correct MC datasetsfrom first data. Therefore, development and tests of the implemented methodare ongoing.



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.008

-0.006

-0.004

-0.002

0

0.002

"Real Data" - MC

MC adjusted - "Real Data"

Figure 6.10: Proof of method using about 5·104 tracks: At the top: a MC distribution (black)of signed transverse track impact parameters is mapped (red) to a pseudo real data distribution(blue). The mapping is done for tracks with 0.1 < η < 1.9 and 1 GeV< pT < 2 GeV for alljet types (b- and light). At the bottom: the differences of the distributions on the left. Thedifference between pseudo real data and MC (blue) represents the deviations of the distributionsbefore mapping. The red curve represents the deviations of the distributions after mapping(adjusted MC − pseudo real data).

6.4 Alternatives 73

6.4 Alternatives

A different approach is explained in [119] and outlined in figure 6.11. This ap-proach was adopted for DØ [120] and is also in progress for ATLAS [116]. There,

Error of IP(MC Distribution)

Error of IP(MC)

Error of IP(Data Distribution)

Error of IP(Data Track Fit)

Error of IP(Data Combined)

IP(MC)

IP (MC)(ADJUSTED)

Scale

Combine

Additive Correction

Figure 6.11: Track impact parameter correction approach of Borisov and Mariotti at DEL-PHI.

the errors of the track impact parameters from the real data distribution andthe track fit are combined first to reflect the real precision/understanding of thetracking system. According to this combined error, the error of the MC distri-bution is corrected. After having corrected the error of the impact parameter,the IP itself can be corrected additively (assuming a gaussian distribution) byusing those errors of MC and real data. Finally, the impact parameters of asubset of tracks are adjusted to correct for the non-gaussian tails of the impactparameter distribution. This approach may be also used to correct data, if theerror calculated by the reconstruction needs some correction.

It is also possible to make a correction on the clusters associated to tracks.This method would provide adjusted track fits with corresponding errors. Itrequires rerunning the reconstruction processing step: the corrections could beapplied to ESDs containing the cluster information or by reprocessing the rawdata. From this adjusted data, AODs can be produced and analyzed afterwards.Therefore, such a procedure may easily take a lot of resources, and more time willbe consumed to do the correction iterations, until a sufficient tuning is reached.Hence, the presented mapping approach may be the preferred way of correctingMC track impact parameters, if the correction performance is similar for bothalternatives.

Chapter 7

Systematic Uncertainties ofAcceptance Corrections

Top quark pair production has a considerable cross section at the LHC (σ(tt) ≈800 pb in proton-proton collisions at design energy). Due to the quite easy se-lection and the sizeable signal to background ratio for the semileptonic channel,the ATLAS detector can be in-situ calibrated by large high purity samples ag-gregated in a few months of data acquisition. Moreover, the top quark propertiesmeasured at the Tevatron or predicted by the Standard Model can be validatedwith high statistics never reached before. Eventually, new physics may be pointedat by analyzing top events.

The precise measurement of top quark properties suffers from limited reso-lution, detection efficiency less than 100% and the presence of background pro-cesses. According to [121], a true distribution ~g (generated) which can be neverexactly measured, of a top quark observable may be estimated from the datadistribution ~d, by subtracting the expected background ~b and applying a multi-plicative correction factor ~C. This has to be done in bins i of the distribution:

gi = Ci · (di − bi) (7.1)

The factors Ci can be determined by a simulation of the experiment. The MonteCarlo program is run once with and once without the detector simulation, yieldingmodel predictions for the observed and true values of each bin, aMC

i (accepted,i.e. generated with reconstruction cuts) and gMC

i . Here ~gMC refers to the signalprocess only, i.e. background is not included. The correction factor is then simplythe ratio

Ci =gMC

i

aMCi

(7.2)

Statistical errors in the correction factors can be made negligibly small by pro-ducing large MC statistics. If, moreover, the effects of the resolution are insignif-icant, an accepted event stays in the same bin as the corresponding generated

76 Chapter 7. Systematic Uncertainties of Acceptance Corrections

one. Thus, the expected number of entries (signal) in bin i without backgroundis the measured difference between accepted and background entries:

asigi = ai − bi =

1

Ci

gi (7.3)

In the ideal case, Ci = 1 and therefore gi = asigi . However, every real experiment

has a bias, that is the deviation from the true distribution, is

si =

(

gMCi

aMCi

− gi

asigi

)

· asigi (7.4)

Since gi is unknown, only a rough estimate of the systematic uncertainty dueto this bias can be obtained by computing the correction factors with differentMonte Carlo models. Equation 7.4 implies that the systematic uncertainty of thecorrection procedure vanishes (si = 0), if the ratio of gi/ai agrees for data andMC (gMC

i /aMCi = gi/a

sigi ). However, this can only be checked by comparing the

data to the accepted MC. If perfect agreement is reached here and if the detectorsimulation represents the experiment, it is justified to assume that the correctionmethod is free of any bias.

Though, forthcoming data may differ from Monte Carlo simulations by morethan the statistical and systematic uncertainties. When new physics or furthersystematic errors are excluded, the deviations of the Monte Carlo simulations mayarise from an imperfect detector simulation or from the physics simulation (MCgenerator). Assuming the correctness of the detector simulation, the averagevariation of the MC generators corresponds to an uncertainty of the physicsmodelling. The deviations have to be reduced by either tuning the generators or,initially, by reweighting their events such that the model events agree with thedata.

The latter approach is chosen and presented in this thesis. The observednumber of top/antitop quark events, which corresponds to their total selectionefficiency when the luminosity is known from measurements, may be different forMC and data. For example, the systematic uncertainty due to the physics modelof MC may be 20%, whereas for data it may be 5% corresponding to detectorand theoretical uncertainties of ≈ 2.5% each. The statistical uncertainty can beneglected because of the high production rate at the LHC and if enough MCevents are available. But if all MC events are reweighted in an appropriate way,the systematic uncertainty of MC is reduced to the level of data. Hence, theagreement between MC and data is improved for the selection efficiency.

Since the integral over a differential distribution dNdx

provides the total numberN ,

N =

∫

dN

dxdx (7.5)

7.1 Monte Carlo Simulation 77

the total numbers agree, if the differential distributions do. However, the fullrange of an observable x is normally not accessible in an experiment and extra-polations have to be done, which may be inaccurate. If these are accurate, theapproach of reweighting depends on whether the differential distributions of theobservables, which are used for selection cuts, are correlated or not. When theyare uncorrelated, the MC events can be reweighted by a total weight, which is theproduct of all weights obtained for the individual distributions. If the observablesare correlated, a reweighting from one distribution may suffice.

The ratio of MC and data provides the inverse of the weight, which may beapplied bin by bin to the MC sample in order to achieve a better agreementwith data. However, such a bin by bin reweighting procedure is unphysical (notsmooth) and leads generally to instabilities. Thus, the distribution can be insta-ble, when the method is applied iteratively. Therefore, the ratio of the MC anddata distribution is fitted by a smooth function, which may be even extrapolatedinto regions that are experimentally not accessible. The fit function can be arbi-trarily chosen, if it is smooth, fits the ratio of the distributions and leads to animprovement of the agreement. The systematic uncertainty due to the choice ofthe fit function can be estimated by using different functions for the reweightingof the MC events.

The next sections explain the simulation of tt-events, the selection of semilep-tonic events and the corresponding efficiencies, the distributions of the relevantobservables, the reweighting approach and its results.

7.1 Monte Carlo Simulation

As explained in Section 2.3, Monte Carlo computer programs also called MCgenerators, that approximate the analytical theory by empirical models, are usedto simulate the measurement of single collision events. In ATLAS, several MCgenerators and many related computer packages, which contain for example PDFsets, are used [122]. Typically, the MC generator output are the four-vectors ofall final state particles of a collision event. However, the Monte Carlo models donot take statistical and systematic uncertainties, which result from the limitedresolution of the measuring device, into account. So, the measured distributionswill be different from the initially predicted distributions.

To reconcile the experiment with the theory, the detector must be simulated.In figure 7.1, all steps of the simulation data processing, also called the ATLASMC full chain, are sketched. The common steps for real data processing are notedas well.The detector simulation gets the predicted observables (i.e. the four-vectors ofall final state particles) from the MC generator and returns a simulated detectorresponse. In ATLAS, the detector simulation GEANT4 [123] is used. The exactdetector material description including geometry and masses is provided to this


Figure 7.1: Program and data flow from MC event generation to analysis (the ATLAS fullchain Monte Carlo production). The flow of fast simulation (ATLFAST) and real data is notedas well. [124]

program. GEANT approximates the exact geometry by a composition of manysimple geometrical objects as cuboids. For each particle provided by the MC gen-erator GEANT produces “hits”, which are energy deposits in sensitive detectorvolumes according to the Bethe stopping power formula or other models.Another computer program, called digitization, converts these energy depositsinto detector responses, “digits”, typically voltages or times on preamplifier out-puts. More details about this digitization step can be found in [125]; the overalldocumentation of the generation and the digitization step, also called simulation,can be read in [126]. The simulated and digitized detector response should pro-vide the same distributions of the experimental observables as the measurement.

Since such a simulation of the ATLAS detector is very time consuming, itmay be appropriate to simulate the detection of particles only in a simplifiedway, which is much faster. Therefrom, the names of the corresponding AT-

7.1 Monte Carlo Simulation 79

LAS packages were chosen: ATLFAST [127] and FATRAS [128], which simulatestracks. To find a balance between speed and consumption of resources on theone hand and accuracy on the other, full and fast simulation can be combinedby ATLFAST2 [124], which may also do the digitization but skip the intensivesimulation step.

Instead of comparing raw detector responses, the MC event is reconstructedjust like a real data event as explained in Chapter 5: Raw data stored in RDOfiles is used to reconstruct common physics objects like tracks, vertices, jets, elec-tron, photons, muons, missing transverse energy. These objects can be savedtogether with some raw information in ESD files, wich are mainly used for de-tector calibration. Physics analyses prefer AOD or DPD files, which are small insize because they contain almost no raw information.

Since the reconstruction algorithms are identical for simulated and real colli-sion events, the distributions of the experimental observables should be the samefor simulated and real data. However, MC simulation can neither predict un-known physics nor reproduce precisely all detector effects. Moreover, various MCgenerators or detector simulations use different models. Therefore, deviationsfrom real data represent the systematic uncertainty caused by the modelling, ifsystematic uncertainties in the data are minimized.

In order to study such uncertainties, 5 MC generators were chosen to simu-late tt-events in proton-proton collisions with a center of mass energy of 14 TeV.Each of the generators mentioned in Section 2.3, Pythia, AcerMC, Sherpa, Her-wig, MC@NLO, simulated about 105 events each. Since this chapter presents amethod, how systematic uncertainties due to the physics modelling can be re-duced by reweighting MC events from (pseudo) data, numbers of events werenot normalized to an integrated luminosity, and any effect of pile-up was neithersimulated nor considered. Though for AcerMC the hadronization step was doneby Pythia and Herwig was used for this purpose for MC@NLO, we stick in thisthesis to the short notations “AcerMC” and “MC@NLO”.

The configuration values of the generators were chosen as default values eitherof the generator for LHC or of ATLAS. After the generators had been configuredin a similar way, for example they used the same top quark mass of 175 GeVor similar PDF sets (CTEQ6L1 or CTEQ6M [129, 130]), they were run withinthe same version of the Athena framework to guarantee a maximal consistency.The detector geometry and simulation as well as the event reconstruction areexactly the same for all 5 generators and defined by the corresponding versionof the Athena environment. All configuration files and scripts, which were usedto generate the 5 datasets [131], can be obtained at [132]. Table 7.1 overviewsall steps of the simulation. In the following section we will explain the eventobservables, selection and analysis.


Step Athena version generator generator version

generation 14.4.0.1

Pythia 6.419Herwig 6.510AcerMC 3.5

MC@NLO 3.31Sherpa 1.1.2.2

detector simulation 14.4.0digitization 14.5.1

reconstruction 14.5.1analysis 14.5.0

Table 7.1: Configuration of 5 Monte Carlo tt simulations. The steps detector simulation toanalysis use the same Athena version and same detector geometry (ATLAS-GEO-02-01-00) forall 5 generators.

7.2 Total and Differential Efficiencies

As mentioned in section 2.2, about 44% of top quark pairs decay semileptonically.Usually, the final state consists of two b-jets and two light jets, one lepton and oneneutrino, apparent as missing transverse energy. High level reconstructed objectssuch as electrons, muons or jets, whose reconstruction was explained in Section6.1, can be defined in several ways. To achieve comparability, the standard defini-tions of ATLAS for top quark physics were chosen [114,133], for example a conejet algorithm with an opening angle of 0.4 rad was used. Since reconstructionalgorithms of different objects run independently over the raw data, the same hitpattern may be reconstructed as more than one object. For example, the samecalorimeter hits can be identified as an electron and as a jet, or generated eventsare crowded by soft Bremsstrahlung photons (and therefrom also by soft elec-tron/positron pairs due to pair production) from electrons, which would be seenaltogether in the detector only as one narrow electromagnetic shower. There-fore, a procedure may be necessary to decide, which of the overlapping objectsare selected and which are removed, hence its name “overlap removal”. In thisanalysis, it was implemented also according to [114,133] and referred as selectioncut 1.Like in [114, 133], the following observables and corresponding cuts due to kine-matics (and for a reduction of background events) are used to select semileptonictt-events:

• one isolated lepton (only e or µ because a τ can decay hadronically) withhigh transverse momentum pT > 20 GeV (selection cut 2)

• large missing transverse energy /ET > 20 GeV (selection cut 3)

7.2 Total and Differential Efficiencies 81

• at least 4 jets with high transverse momentum pT > 20 GeV (selectioncut 4)

• of which at least 3 have an even larger transverse momentum pT > 40 GeV(selection cut 5)

Further requirements can be that 2 jets have an invariant mass inside a windowaround the W boson mass or there are one or two b-tagged jets (only useful forlater data when all detector effects are fully understood). Although jets can bemeasured in the range of the pseudorapidity |η| < 3.2 by the ECAL/HCAL oreven up to |η| < 4.9 by the FCAL, b-tagging is possible only in the range |η| < 2.5,which corresponds to the acceptance of the inner detector. Therefore, only thoseevents are selected, whose aforesaid objects (electrons, muons, jets) are all within|η| < 2.5. Since this study is a first approach, no trigger information are used.The resulting selection efficiencies of the 5 cuts above for events without (genera-tor level) and with detector simulation (reconstruction level) out of a semileptonicsample are illustrated in table 7.2. For a certain MC generator and selection cut,the relative deviations between generator and reconstruction level are up to 13%.For example, by the fourth cut (≥ 4 jets with pT > 20 GeV) 78% (19.1/24.5) ofthe events on generator level are kept, whereas 68% (19.4/28.5) are kept on re-construction level, corresponding to a relative deviation of 13% for Herwig. Suchdeviations may be explained by the acceptance and resolution effects of the de-tector simulation and the fact, that there is for example no precise counterpart ofa jet in both levels. Likewise, the large acceptance differences between generatorand reconstruction level in the first cut are caused by soft particles, which areremoved on generator level.Similarly, the relative deviations between two MC generators for the same selec-tion cut represent up to 11%, which can be explained by the different modellingof the physics process by the generators. The maximal deviation between thetotal efficiencies is 20% (12.7/15.8) for Pythia versus Sherpa after the fifth cut onreconstruction level. This difference is mainly due to Pythia’s new parton showeralgorithm, which is much more radiative. The statistical and systematic uncer-tainty for data will probably be much smaller (. 5%). Therefore, a reweightingof MC events is already necessary from the point of view of total efficiencies.

The presumably most precise reconstruction of a semileptonic tt-event is akinematic fit. As indicated in Section 2.2, two light jets must come from a Wdecay and have a corresponding invariant mass. Similarly, the missing transverseenergy and a charged lepton have to come from another W decay. The charge ofthe lepton determines, if it comes from the top or antitop quark. Furthermore,two jets have to be b-tagged. Each of them must have together with one of thereconstructed W bosons an invariant mass of a top quark. Such a fit may bechallenging, especially for early data, when the detector is not well understood.Hence, this reconstruction strategy is not chosen in this analysis. Since thereconstruction of the leptonically decaying top quark depends on three or four


different object types (missing transverse energy, jet, electron or muon), it isneglected here. Especially the missing transverse energy will be not well measured

Gen. Cut [%] Pythia Herwig AcerMC MCNLO Sherpa

none 100 100 100 100 100no. of events (≡ 38348) (≡ 40658) (≡ 41777) (≡ 28198) (≡ 96711)stat. error ±0.2 ±0.2 ±0.2 ±0.2 ±0.1overlap

91.8 89.8 91.1 90.6 90.0removalisol. lepton

25.8 25.7 25.9 25.8 26.2pT > 20 GeV/ET > 20 GeV 24.5 24.5 24.6 24.6 24.7≥ 4 jets

20.2 19.1 20.1 19.4 18.9pT > 20 GeV≥ 3 jets

15.2 13.4 15.0 13.6 12.9pT > 40 GeV

Reco. Cut [%] Pythia Herwig AcerMC MCNLO Sherpa

none 100 100 100 100 100no. of events (≡ 38345) (≡ 40648) (≡ 41758) (≡ 28188) (≡ 96677)stat. error ±0.2 ±0.2 ±0.2 ±0.2 ±0.1overlap

99.9 99.9 99.9 99.9 99.9removalisol. lepton

29.7 31.0 30.0 30.5 30.5pT > 20 GeV/ET > 20 GeV 27.3 28.5 27.6 27.9 27.8≥ 4 jets

20.1 19.4 20.0 19.6 18.1pT > 20 GeV≥ 3 jets

15.8 14.4 15.7 14.6 12.8pT > 40 GeV

Table 7.2: Event selection efficiencies out of semileptonic samples of five MC generators. Aswell, the numbers of events and the corresponding statistical error are noted, which applies ap-proximately to the efficiencies of a column. At the top: without detector simulation (generatorlevel). At the bottom: with detector simulation (reconstruction level).

in early data. Whereas, the hadronically decaying top quark relies only on theobject type jet (1 b-jet and 2 light jets from the W decay). According to [114],we define this top quark as that combination of three jets (out of N jets, hencethere are N !

3!·(N−3)!possible combinations), which has the highest vector sum of

their transverse momenta:3

∑

j=1

~pTj= max (7.6)


However, this loose definition selects quite probably a wrong jet combination nearthe threshold of tt pair production and neglects b-tagging and other constraints.In spite of the corresponding combinatorial background, this simple definitionmay be sufficient for first data and represent a stable analysis.

For the differential distributions of the top quark the expected resolutions weredetermined from two dimensional plots: The difference of the reconstructed andgenerated observable is plotted against the generated observable. An example ofa resolution plot for the transverse momentum of the top quark is illustrated infigure 7.2. The binning of the plots of the differential cross sections was chosen a

T,gen. - p

T,reco.p

-200 -100 0 100 200 300 400

3

Ent

ries

[1]

0

100

200

300

400

500

600

Figure 7.2: Example of a resolution plot: The expected resolution (reconstructed – generated)of the transverse momentum of the hadronically decaying top quark for Pythia in the range of200 to 220 GeV (generated) is about 13 GeV.

bit coarser and more regular than the obtained resolutions, see figure 7.3. Thisbinning ensures that bin by bin migration between generated and accepted eventsin the MC simulation are small.Since LHC and ATLAS had not collected any collision data during these studies,Sherpa was arbitrarily chosen to represent pseudo data. As patent in figure7.4, relative to pseudo real data there are deviations only up to 5% in bins of thepseudorapidity of the hadronically decaying top quark. Hence the MC generatorsagree quite well in this observable, even with detector simulation. However, inbins of the transverse momentum there are deviations up to a factor of 3, see figure7.5. Due to this discrepancies, a reweighting of the MC events, for example fromthis observable, is mandatory.


Figure 7.3: The expected resolutions (reconstructed – generated) of the transverse momentumand pseudorapidity of the hadronically decaying top quark for Pythia.


Eta [1]

0 0.5 1 1.5 2 2.5

ηddn n1

0

0.01

0.02

0.03

0.04

0.05

Pythia

Herwig

AcerMC+Pythia

MCNLO+Herwig

Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa

AcerMC+Pythia/Sherpa

MCNLO+Herwig/Sherpa

Figure 7.4: Distributions of the pseudorapidity of the hadronically decaying top quark ongenerator level without any cuts. On reconstruction level (with cuts) the plots are very similar,see B.3. At the top: for five Monte Carlo generators. At the bottom: four of them relative tothe last (Sherpa arbitrarily chosen as pseudo data).

86C

hap

ter7.

System

aticU

ncertain

tiesof

Accep

tance

Correction

s

Transverse Momentum [GeV]

100 200 300 400 500 600 700 800 900 1000

Tdpdn n1

-310

-210

-110

Pythia

Herwig

AcerMC+Pythia

MCNLO+Herwig

Sherpa


100 200 300 400 500 600 700 800 900 1000

Tdpdn n1

-410

-310

-210

-110

Pythia

Herwig

AcerMC+Pythia

MCNLO+Herwig

Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

MC

/"R

D"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure 7.5: Distributions of the transverse momentum of the hadronically decaying top quark for five Monte Carlo generators without detectorsimulation (generator level without any cuts) at the left, with detector simulation (reconstruction level with cuts) at the right. at the bottom.The corresponding ratio plots are shown as well (Sherpa arbitrarily chosen as pseudo data).

7.3 Reweighting of MC Events from (Pseudo) Data 87

7.3 Reweighting of MC Events from (Pseudo)

Data

As announced in the previous sections, a reweighting of the MC events from(pseudo) data may be appropriate and sufficient to reduce the uncertainty of thephysics modelling (several MC generators with different predictions for the sameprocess). Though, it is quite unprobable to obtain good agreement between MCand (pseudo) data for the total and all relevant differential efficiencies. This maybe even so after the MC events have been reweighted. The main reasons arelimited experimental resolution, noise and the fact, that only parts of the fullmomentum and rapidity ranges can be measured with ATLAS. Particularly, dis-tributions have to be extrapolated into the low energy/momentum range, wheremeasurements are not possible. However, if the MC generators have a similarapproach for the important stages of event generation (explained in Section 2.3),and differ only in one specific domain or variable, it is likely that the differentMC models agree reasonably well after the reweighting according to a single, wellchosen observable.

In fact, the above situation is realized. There are three major elements inthe generation of a tt event: the underlying PDFs, the hard strong interactiontt-production process, and the weak W decay. The PDFs are by constructionsimilar in the performed simulation, and should mainly influence the longitudinalproperties (η distribution) of the tt event. The decay of the W is theoreticallyvery well understood and mis-simulation is unlikely. The largest differences areexpected for the strong production process, already due to the differing values ofthe strong coupling and the ansatz (matrix element, parton shower, matching ofboth) of the models.

Variations due to the hadronization model should be of minor significance,as the mass scales of the involved processes — the top mass with respect totypical hadron masses — are substantially different. Thus, an observable forthe reweighting is suggested, which shows large differences between the modelsand is sensitive to the strong production process. Therefore, the hadronicallydecaying top quark was chosen, which is obviously sensitive to the strong pro-duction process and is defined according to formula 7.6. Since the distributions ofthe pseudorapidity and the azimuthal angle for that observable agree quite wellbetween (pseudo) data and MC, they are not considered in this first approach.However, the transverse momentum of the top quark differs up to a factor of 3from pseudo data (Sherpa arbitrarily chosen) and is therefore the obvious candi-date.

The reweighting procedure is neither stable nor physical, when the MC eventsare reweighted bin by bin from the ratio of the MC and pseudo data distribution(this implies detector simulation and all selection cuts mentioned in Section 7.1and 7.2). Therefore, it is fitted for each MC generator. A combination of 2


straight lines (y1,2 = a1,2 · x1,2 + b1,2 for two x = pT ranges 1 and 2) was chosen,representing the most simple smooth function describing the ratio. Figure 7.6illustrates such fits for Herwig. The corresponding figures for the other MCgenerators are in Appendix B.2. Some data points do not lie on the fit function,


100 200 300 400 500 600 700 800 900 1000

MC

/"R

D"

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Figure 7.6: The fits to the distribution of the transverse momentum of the hadronicallydecaying top quark for Herwig relative to pseudo data (Sherpa). The fit parameters are chosenso that the overall function is smooth.

especially at low transverse momentum the fit function lies beneath the data.Though an extrapolation to values near zero may lead to bad results for thisreason, a reweighting was tried.The weight, which has to be given to the MC events, is provided by followingformula, where a and b are coefficients from the straight line fit and pMC,0c

T isthe transverse momentum of the hadronically decaying top quark (as defined byequation 7.6) without detector simulation and without any cuts:

w =(

a · pMC,0cT + b

)−1

(7.7)

If the MC event has already a weight (for example MC@NLO produces weightedevents), the total new weight is just the product of the old and the calculatedweight.

After all MC events had be reweighted, the ratio between MC and (pseudo)data is nearly flat and as expected about one for the transverse momentum ofthe hadronically decaying top quark. Figure 7.7 illustrates the improvement withand without detector simulation. This shows that the chosen simple fit strategyworks well.



100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure 7.7: Reweighted distributions of the transverse momentum of the hadronically de-caying top quark for four Monte Carlo generators relative to pseudo data, without detectorsimulation (generator level) and any selection cuts at the top, with detector simulation (recon-struction level) and selection cuts at the bottom.

Especially at the connection point of both straight lines, where the fit functionis not differentiable, the reweighting provides not worse results than in otherranges. Moreover, the extrapolation in the low momentum region seems not tofail, because the distribution with detector simulation is still below pseudo data.Though the MC events were strongly weighted up in that range, what can be seenin the distributions without detector simulation, a migration of reconstructedevents to higher bins is apparently unprobable.

Moreover, the distribution of the pseudorapidity of the hadronically decayingtop quark does not change by the reweighting, as it is independent of the trans-verse momentum. Finally, all distributions of observables serving as selection cutsshow for each bin deviations between MC and pseudo data of about the same


size as before reweighting or even up to a factor of 10 smaller. The correspondingfigures are patent in the appendix B.2. The improvement is very pronounced forthe transverse momentum of the jets (figure B.10), especially the leading fourjets (figures B.11, B.12, B.13, B.14). This is expected due to the definition ofthe hadronically decaying top quark by the transverse momentum of three jets,see formula 7.6. An improvement can be also seen for the reconstructed missingtransverse energy (figure B.8). Since the transverse energy is mainly calculatedfrom the calorimeter signal after a subtraction of the jets, an improvement ofthat observable is reasonable.

However, the deviations for the pT spectrum of electrons and muons (figuresB.4 and B.6) stay in the same range. As high energetic electrons/muons orig-inate from the leptonically decaying top quark, there may be only a reducedcorrelation to the transverse momentum of the hadronically decaying top quark,and hence no improvement for those distributions is visible. Similarly to thehadronically decaying top quark, the deviations between the distributions of thepseudorapidities of all object types remain as before reweighting.

Overall, the observed behavior is consistent with the conjecture on page 88that the differences between the models are predominantly due to differences inthe hard production process. The same is to be expected for comparisons of themodels to real data as weak W decays are very well understood and longitudinalproperties of top production seem less important in view of the reconstructionefficiencies.

To analyze the improvement of the total efficiency, we define it as the numberof reconstructed data events accepted after N selection cuts (DNc) divided by thenumber of generated (no detector simulation) events without any cuts (MC0c):

ǫ =DNc

MC0c(7.8)

Table 7.3 illustrates the resulting improvements gained by the reweighting. Forexample, the deviation of the total efficiency of Herwig from pseudo data dropsfrom +13% to 0%. The mean improvement is obtained by averaging over thefour generators: the deviation of this average generator from pseudo data dropsfrom ≈ +18% to ≈ +3%. The mean deviation over the four generators is ≈ +3(= (+4 + 0 + 6 + 3)/4). Therefore, the aimed uncertainty smaller than 5% isreached, except other models (generators) deviate even more than the chosenfour. However, another generator that deviates even stronglier than Pythia orHerwig, which are very mature, should not be taken too serious.

Hence, there is a quite unexpectedly large improvement for the total andmany differential efficiencies by such a simple reweighting approach using onlythe distribution of one observable. This may be explained as above. However, theimprovement may depend on the extrapolation into the low momentum range.

In order to estimate the error of the approach, the reweighting is tried byusing the transverse energy HT of the tt pair, represented by the four leading


DNc

MC0c [%] Pythia Herwig AcerMC MCNLO av. Gen Sherpa

no. of events 35070 36412 37837 24412 33433 86989stat. err. ±0.2 ±0.2 ±0.2 ±0.2 ±0.2 ±0.1unweigh. 15.8 14.4 15.7 14.6 15.1 12.8rel. to S. +23 +13 +23 +14 +18 0reweigh. 13.3 12.8 13.6 13.2 13.2 12.8rel. to S. +4 0 +6 +3 +3 0

Table 7.3: Total selection efficiencies of semileptonic tt events before and after reweightingfrom pseudo data (Sherpa) for four Monte Carlo generators. Their relative deviations frompseudo data are given as well. The sixth column represents the average generator of these fourand its deviation from pseudo data (not the average deviation of the four generators). Thecorresponding statistical errors are also noted, which apply approximately to the efficiencies ofa column.

jets, the charged lepton and the missing transverse energy. Since the reweightingon the basis of the transverse momentum of the hadronically decaying top quarkshows no improvement for the differential efficiencies of the transverse momentumof muons and electrons, a reweighting on the basis of HT seems a priori to bedoomed. The corresponding fit for Herwig, again with two straight lines, isillustrated in figure 7.8 and looks not better than the one in figure 7.6.

[MeV]TH

200 400 600 800 1000 1200 1400

MC

/"R

D"

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Figure 7.8: The fits to the distribution of the transverse energy HT for Herwig relative topseudo data (Sherpa). The fit parameters are chosen so that the overall function is smooth.


The obtained results are similar to those of the standard approach: as illus-trated in figure 7.9, the reconstructed HT (with detector simulation and cuts) is

Et [MeV]

200 400 600 800 1000 1200 1400

MC

/"R

D"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Et [MeV]

200 400 600 800 1000 1200 1400

MC

/"R

D"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure 7.9: Original (at the top) and reweighted (at the bottom) distributions of the trans-verse energy HT for four Monte Carlo generators relative to pseudo data, with detector simu-lation (reconstruction level) and selection cuts.

quite flat and near 1 after the reweighting procedure. For the total efficiencieseven a slightly better improvement is obtained. As noted in table 7.4, the devi-ation of the average generator from pseudo data, which is obtained by averagingover the four generators, drops from ≈ +18% to ≈ −1%. The mean deviationover the four generators is ≈ −1% (= (0 − 4 − 2 + 3)/4).

Therefore, the estimated obtainable systematic uncertainty of ≈ 3% is reli-able, and the first reweighting approach looks very promising to reduce potentialdeviations of MC from first data.

7.4 Alternatives 93

DNc

MC0c [%] Pythia Herwig AcerMC MCNLO av. Gen Sherpa

no. of events 35070 36412 37837 24412 33433 86989stat. err. ±0.2 ±0.2 ±0.2 ±0.2 ±0.2 ±0.1unweigh. 15.8 14.5 15.8 14.6 15.2 12.9rel. to S. +22 +12 +22 +13 +18 0reweigh. 12.9 12.4 12.7 13.3 12.8 12.9rel. to S. 0 −4 −2 +3 −1 0

Table 7.4: Total selection efficiencies of semileptonic tt events before and after reweightingfrom HT of pseudo data (Sherpa) for four Monte Carlo generators. Their relative deviationsfrom pseudo data are given as well. The sixth column represents the average generator of thesefour and its deviation from pseudo data (not the average deviation of the four generators). Thecorresponding statistical errors are also noted, which apply approximately to the efficiencies ofa column.

7.4 Alternatives

A different approach was suggested to estimate the systematic uncertainty of thetotal selection efficiencies of top/antitop quark events: a chi-square method. Thechi-square of an observable is defined as the ratio between the squared deviationof MC from data RD and the sum of their squared errors. For a distribution withbins i, this has to be summed over all bins N :

χ2 =N

∑

i

(MCi − RDi)2

σ2MC,i + σ2

RD,i

(7.9)

If the chi-square is calculated for several MC simulations, an estimate of thesystematic uncertainty can be obtained by fitting a plot of the chi-square of theobservable in dependence of the observable: The distance between the position ofthe RD entry (data has χ2 = 0) and the position of the minimum of the parabola,which is fitted to the MC values and forced to be also at χ2 = 0.

Such a chi-square method is usually applied, for example, for one MC gen-erator: a chosen variable of the MC generator, for example the top quark mass,is varied and after the simulation different distributions of an observable areobtained.

Nevertheless, the total event selection efficiencies (reconstruction level) takenfrom table 7.2 was chosen as an observable (N = 1), the chi-square was calculatedfor the four MC generators (“overall” chi-square) and Sherpa served as pseudo realdata like above. As a second try, the distributions of the 10 selection observablesof Section 7.2 (η of electron, muon, jet; /ET; pT of electron, muon, 1st, 2nd,3rd, 4th leading jet) were chosen (Nj = number of bins for each distributionj). The sum over the computed chi-squares was calculated, which, however,


neglects correlations between the 10 observables. The chi-squares of both triesare illustrated for all MC generators in figure 7.10. Parabolas (a · x2 + b · x + c)

Figure 7.10: Chi-square in dependence of the total selection efficiency for five MC generators.The red symbols correspond to the overall chi-square, the blue symbols correspond to the sumover 11 chi-squares. In order to have a comparability between both, the overall chi-square ismultiplied by 11. Solid lines indicate a parabolic fit to 4 MC generators, dashed lines representsuch a fit to 5 MC generators. For a visibility of both Sherpa entries the red one was displacedby +0.1.

are fitted to both chi-squares. However, these fits have few significance, becausethe entries of MCNLO and Herwig and of Pythia and AcerMC are almost ontop of each other. In order to obtain significant results, more MC generators areneeded.

Furthermore, the pseudo real data entry lies outside the entries of the MCgenerators. Therefore, the estimate of the systematic uncertainty with cor-responds at least to the distance to the next entry, which results in ≈ 13%(= 100 · (14.4− 12.7)/12.7). This value is almost by a factor of 4 larger than thevalue obtained by the reweighting approach in the previous section.

7.4 Alternatives 95

Eventually, first data is between the entries of the MC generators, i.e. in theminimum of the dashed parabolas (about 13.0). Only then, the systematic un-certainty would be at the 1% level, and the chi-square method would be preferredto the reweighting approach.

Chapter 8

Summary and Outlook

The LHC and its experiments will open new possibilities to study precisely theproperties of known particles such as the top quark and to find particles neverobserved before.

Careful operation, maintenance and tuning of the detectors and their readoutsystems are mandatory, in order to obtain high statistics data samples and per-form precision measurements. The control and the readout of the ATLAS pixeldetector is performed through an optical transmission line. The installation andcommissioning of the optical cables as well as the steps from the production tothe operation of the Back of Crate cards, which are the optical interface on theoff-detector side, are presented in this thesis. Tests of the optical cables have beendeveloped and realized. These tests ensure a complete functionality of all opticalcables and prove its operation within given limits. Similarly, the full operabilityof the Back of Crate cards is assured by a precise production, extensive tests anda careful installation. Since spring 2008, the readout of the pixel detector is readyfor operation. Data was taken successfully during cosmics runs, and the tuningof detector and readout is ongoing.

Track impact parameters, which are mainly determined by the pixel detector,are important observables for b-tagging and therefore for many physics analysessuch as top quark studies. Systematic uncertainties are reduced, if distributionsof MC simulation and data agree. A method to map MC distributions of trackimpact parameters on data was developed, implemented and tested by usingmisaligned MC samples as data. The performance of the approach is promisingfor the correction of MC samples from first data. Especially for impact parameters|IP | < 0.4 mm the agreement between MC and pseudo data was improved by upto a factor 4. The work to prepare the corresponding software package for firstdata is ongoing as well as tests with current MC data samples.

Precision measurements of observables of top quark pair events will be done byATLAS using the LHC as a top factory. From the experience gained by the Teva-tron experiments CDF and DØ Monte Carlo generators were tuned to predict topquark observables at the LHC. A comparison of five common MC generators was

98 Chapter 8. Summary and Outlook

done for the total selection efficiencies and for many distributions, for examplethe transverse momentum and pseudorapidity of the hadronically decaying topquark. The deviations between the total efficiencies of the MC generators are onaverage 18%, the maximum is 23%. The deviations between the MC distributionsof the observables serving for the event selection depend on the observable. Forthe pseudorapidity the deviations are typically small (≈ 2%). However, there arelarge deviations for the transverse momenta, especially of jets and the hadroni-cally decaying top quark (up to a factor of 4 in certain bins). These deviationsrepresent the systematic uncertainty caused by the physics modelling.Likewise, detector effects on these distributions were studied by using a full de-tector simulation and event reconstruction. As expected, the distortions of thedistributions by detector effects are marginal or linear.A method was developed, implemented and tested to reweight the MC distribu-tions from one observable, for example the transverse momentum of the hadroni-cally decaying top quark, such that they agree better with pseudo data. Thereby,an improvement of the average systematic uncertainty of the total efficiencies bya factor of > 5 to . 3% was reached. This was confirmed by using a differentobservable for reweighting. Such a performance may be also obtained by apply-ing this method on first data. Therefore, the work to prepare the correspondingsoftware package is ongoing.

Appendix A

Implementation and Plots of theTrack Impact Parameter Tuning

A.1 Implementation of the Track Impact Pa-

rameter Mapping Approach

A new ATLAS offline software package ImpactParameterTuning was created toadjust the track impact parameters of Monte Carlo simulation Analysis ObjectData (AOD) files. The package is implemented as a C++ library mostly accordingto the coding rules [134–137]. It can be downloaded at [138]. The necessaryalgorithms can be run inside the Athena framework [66] by using a python joboption file.

The package is split into parts:

• “TrackSelector” is an Athena algorithm, which runs over AODs (MonteCarlo or Real Data — so, a comparison of simulated and real data canbe easily done) and stores event/jet/track information according to chosenselection criteria (for example optimized for tt events) in a ROOT [139] file.

• “TrackIpPlotFitter” is an Athena algorithm, which runs over the .root filegenerated by “TrackSelector” and creates distribution histograms of signedimpact parameters and significances. These histograms may be fitted by aprovided function.

• “TrackAdjustor” is an Athena algorithm, which runs over MC AODs, copiestheir content into new AODs and adds a newRec::TrackParticleContainer. This contains the copied tracks whosetrack impact parameters and error matrix were adjusted.

• “TrackAdjustorTool” is an Athena tool, which is called by “TrackAdjustor”.It creates a new Rec::TrackParticle (with the adjusted track impact pa-

100 Chapter A. Implementation/Plots of the Track Impact Parameter Tuning

rameters) by using the histograms of “TrackIpPlotFitter”, which are pro-vided by two .root files (one for the MC distributions, one for the real datadistributions) in the current directory. The “TrackAdjustorTool” uses twofurther subtools: “TrackParametersAdjust”, which adjusts the track im-pact parameters, and “TrackErrorMatrixAdjust”, which adjusts the trackerror matrix.

Figure A.1 illustrates the program and data flow. More detailed documentationof these parts can be found in their doc directories.

Event Data File(MC AOD)

Track Data File(MC Ntuple)

MC Track IPHistogram

Event Data File(Data AOD)

Track Data File(Data Ntuple)

Data Track IPHistogram

Event Data File(MC AOD)

Event Data File(ADJUSTED)

TrackSelector

TrackIpPlotFitter

TrackAdjustorFigure A.1: Program and data flow of the track impact parameter tuning package.

A.2 Plots of the Track Impact Parameter Mapping Approach 101

A.2 Plots of the Track Impact Parameter Map-

ping Approach

The distributions of the signed transverse and longitudinal impact parameter areshown for MC (perfect detector geometry), pseudo data (misaligned MC) andfor MC after tuning. There are 4 ranges in the transverse momentum of thetrack: 1 − 2 GeV, 2 − 3 GeV, 3 − 4 GeV, > 4 GeV. For the pseudorapidity of thetrack 5 ranges are chosen: (−2.5;−1.9), (−1.9;−0.1), (−0.1; +0.1), (+0.1; +1.9),(+1.9; +2.5). Hence, there are 20 plots for each signed impact parameter (d0 andz0).

102C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.03

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

"Real Data" - MC


Figure A.2: At the top: a MC distribution (black) of signed transverse track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with −0.1 < η < 0.1 and 1 GeV < pT < 2 GeV (at the left) and 2 GeV < pT < 3 GeV (at the right).

A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

103


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.04

-0.03

-0.02

-0.01

0

0.01

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

"Real Data" - MC


Figure A.3: At the top: a MC distribution (black) of signed transverse track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with −0.1 < η < 0.1 and 3 GeV < pT < 4 GeV (at the left) and 4 GeV < pT (at the right).

104C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.01

-0.008

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.008

-0.006

-0.004

-0.002

0

0.002

"Real Data" - MC


Figure A.4: At the top: a MC distribution (black) of signed transverse track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with 1 GeV < pT < 2 GeV and −1.9 < η < −0.1 (at the left) and 0.1 < η < 1.9 (at the right).

A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

105


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-510

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.02

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



106C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-510

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-510

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.03

-0.02

-0.01

0

0.01

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.035

-0.03

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

"Real Data" - MC



A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

107


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-510

-410

-310

-210

-110

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

"Real Data" - MC


Figure A.7: At the top: a MC distribution (black) of signed transverse track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with 4 GeV < pT and −1.9 < η < −0.1 (at the left) and 0.1 < η < 1.9 (at the right).

108C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.004

-0.002

0

0.002

0.004

0.006

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.008

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC



A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

109


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



110C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.03

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.015

-0.01

-0.005

-0

0.005

0.01

"Real Data" - MC



A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

111


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.08

-0.06

-0.04

-0.02

0

0.02

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

"Real Data" - MC


Figure A.11: At the top: a MC distribution (black) of signed transverse track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with 4 GeV < pT and −2.5 < η < −1.9 (at the left) and 1.9 < η < 2.5 (at the right).

112C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g

Signed Longitudinal Impact Parameter [mm]

-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.004

-0.002

0

0.002

0.004

0.006

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC


Figure A.12: At the top: a MC distribution (black) of signed longitudinal track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with −0.1 < η < 0.1 and 1 GeV < pT < 2 GeV (at the left) and 2 GeV < pT < 3 GeV (at the right).

A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

113


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.004

-0.002

0

0.002

0.004

0.006

0.008

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.003

-0.002

-0.001

0

0.001

0.002

0.003

0.004

"Real Data" - MC


Figure A.13: At the top: a MC distribution (black) of signed longitudinal track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with −0.1 < η < 0.1 and 3 GeV < pT < 4 GeV (at the left) and 4 GeV < pT (at the right).

114C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.004

-0.003

-0.002

-0.001

0

0.001

0.002

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.003

-0.002

-0.001

0

0.001

0.002

"Real Data" - MC


Figure A.14: At the top: a MC distribution (black) of signed longitudinal track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with 1 GeV < pT < 2 GeV and −1.9 < η < −0.1 (at the left) and 0.1 < η < 1.9 (at the right).

A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

115


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.005

-0.004

-0.003

-0.002

-0.001

0

0.001

0.002

0.003

"Real Data" - MC



116C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.008

-0.006

-0.004

-0.002

0

0.002

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.008

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC



A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

117


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.01

-0.008

-0.006

-0.004

-0.002

0

0.002

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.01

-0.008

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC


Figure A.17: At the top: a MC distribution (black) of signed longitudinal track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with 4 GeV < pT and −1.9 < η < −0.1 (at the left) and 0.1 < η < 1.9 (at the right).

118C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.006

-0.004

-0.002

0

0.002

0.004

"Real Data" - MC



A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

119


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.008

-0.006

-0.004

-0.002

0

0.002

0.004

0.006

"Real Data" - MC



120C

hap

terA

.Im

plem

entation

/Plots

ofth

eTrack

Impact

Param

eterTunin

g


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110 MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.02

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.02

-0.015

-0.01

-0.005

0

0.005

"Real Data" - MC



A.2

Plots

ofth

eTrack

Impact

Param

eterM

appin

gA

pproach

121


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-410

-310

-210

-110MC adjusted

MC

"Real Data"


-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

"Real Data" - MC



-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

Arb

itrar

y un

its

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

"Real Data" - MC


Figure A.21: At the top: a MC distribution (black) of signed longitudinal track impact parameters is mapped (red) to a pseudo real datadistribution (blue). At the bottom: the differences of the distributions at the top. The difference between pseudo real data and MC (blue)represents the deviations of the distributions before mapping. The red curve represents the deviations of the distributions after mapping.The mapping is done for tracks with 4 GeV < pT and −2.5 < η < −1.9 (at the left) and 1.9 < η < 2.5 (at the right).

122 Chapter A. Implementation/Plots of the Track Impact Parameter Tuning

Appendix B

Implementation and Plots of theReweighting of MC top/antitopEvents

B.1 Implementation of the Analysis and of the

Reweighting

A new ATLAS offline software package DifferentialCrossSectionTTbar was cre-ated to analyze Monte Carlo simulation Analysis Object Data (AOD) files. Thepackage is implemented as a C++ library mostly according to the coding rules[134–137]. It can be downloaded at [140]. The necessary algorithms can be runinside the Athena framework [66] by using a python job option file.

The package is split into parts:

• “EventSelectorTTbar” is an Athena algorithm, which runs over AODs (MonteCarlo or Real Data — so, a comparison of simulated and real data can beeasily done) and stores histograms of all relevant observables stepwise ac-cording to chosen selection criteria in a ROOT [139] file. The functions forreweighting MC events are provided by a python job option file.

• “MCPlotter” is a standalone program, which runs over the .root files createdby the algorithm above, normalizes the histograms and saves them in a.root, .pdf, .ps or .eps file. Fits of distributions for reweighting of MCevents are done manually in the first approach.

More detailed documentation of these parts can be found in the doc directory ofthe package.

124 Chapter B. Implementation/Plots of the Reweighting of top pair Events

B.2 Plots of the Analysis and of the Reweight-

ing

The fits to the ratios between MC and pseudo data (Sherpa arbitrarily chosen)for the transverse momentum of the hadronically decaying top quark (as definedby formula 7.6) are provided for four common MC generators. Furthermore,these ratios are plotted before and after reweighting of the MC events by usingthose fit functions. The plots are given with detector simulation and cuts orwithout detector simulation and without any cuts. Finally, the distributions ofthe observables used for cuts are shown before and after reweighting, both withdetector simulation and cuts or without detector simulation and without anycuts: the transverse momentum of electrons, muon, jets (and the leading 4 jets),the missing transverse energy, the pseudorapidity of electrons, muons, jets.


100 200 300 400 500 600 700 800 900 10003

M

C/"

RD

"

1

2

3

4

5


100 200 300 400 500 600 700 800 900 1000

MC

/"R

D"

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6


100 200 300 400 500 600 700 800 900 10003

MC

/"R

D"

00.20.40.60.8

11.21.41.61.8

22.22.4


100 200 300 400 500 600 700 800 900 10003

M

C/"

RD

"

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Figure B.1: The ratio between MC and pseudo data (Sherpa arbitrarily chosen) for thetransverse momentum of the hadronically decaying top quark in semileptonic tt events forfour common Monte Carlo generators. Blue: Pythia, green: Herwig, brown: AcerMC, grey:MC@NLO. The obtained smooth fit functions are drawn as well.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g125


100 200 300 400 500 600 700 800 900 1000

MC

/"R

D"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.2: Transverse momentum of the hadronically decaying top quark in semileptonic tt events for four common Monte Carlo generatorsrelative to pseudo data (Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts.At the left: before reweighting, at the right: after reweighting.

126C

hap

terB

.Im

plem

entation

/Plots

ofth

eR

eweigh

ting

oftop

pair

Even

ts

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

M

C/"

RD

"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.3: Pseudorapidity of the hadronically decaying top quark in semileptonic tt events for four common Monte Carlo generators relativeto pseudo data (Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At theleft: before reweighting, at the right: after reweighting.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g127


50 100 150 200 250 300 350 400

MC

/"R

D"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

MC

/"R

D"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

MC

/"R

D"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

MC

/"R

D"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.4: Transverse momentum of electrons in semileptonic tt events for four common Monte Carlo generators relative to pseudo data(Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: beforereweighting, at the right: after reweighting.

128C

hap

terB

.Im

plem

entation

/Plots

ofth

eR

eweigh

ting

oftop

pair

Even

ts

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.5: Pseudrapidity of electrons in semileptonic tt events for four common Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: before reweighting,at the right: after reweighting.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g129


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.6: Transverse momentum of muons in semileptonic tt events for four common Monte Carlo generators relative to pseudo data(Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: beforereweighting, at the right: after reweighting.

130C

hap

terB

.Im

plem

entation

/Plots

ofth

eR

eweigh

ting

oftop

pair

Even

ts

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

M

C/"

RD

"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

M

C/"

RD

"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

M

C/"

RD

"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5

M

C/"

RD

"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.7: Pseudrapidity of muons in semileptonic tt events for four common Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: before reweighting,at the right: after reweighting.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g131

E [GeV]

100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

E [GeV]

100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

E [GeV]

100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

E [GeV]

100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.8: Missing transverse energy in semileptonic tt events for four common Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: before reweighting,at the right: after reweighting.

132C

hap

terB

.Im

plem

entation

/Plots

ofth

eR

eweigh

ting

oftop

pair

Even

ts

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Eta [1]

0 0.5 1 1.5 2 2.5

MC

/"R

D"

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.9: Pseudrapidity of jets in semileptonic tt events for four common Monte Carlo generators relative to pseudo data (Shepra arbitrarilychosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: before reweighting, at theright: after reweighting.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g133


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.10: Transverse momentum of jets in semileptonic tt events for four common Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At the left: before reweighting,at the right: after reweighting.

134C

hap

terB

.Im

plem

entation

/Plots

ofth

eR

eweigh

ting

oftop

pair

Even

ts


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.11: Transverse momentum of the first leading jet in semileptonic tt events for four common Monte Carlo generators relative topseudo data (Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At theleft: before reweighting, at the right: after reweighting.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g135


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


100 200 300 400 500 600 700 800 900 1000

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.12: Transverse momentum of the second leading jet in semileptonic tt events for four common Monte Carlo generators relative topseudo data (Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At theleft: before reweighting, at the right: after reweighting.

136C

hap

terB

.Im

plem

entation

/Plots

ofth

eR

eweigh

ting

oftop

pair

Even

ts


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.13: Transverse momentum of the third leading jet in semileptonic tt events for four common Monte Carlo generators relative topseudo data (Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At theleft: before reweighting, at the right: after reweighting.

B.2

Plots

ofth

eA

naly

sisan

dof

the

Rew

eightin

g137


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa


50 100 150 200 250 300 350 400

M

C/"

RD

"

0.5

1

1.5

2

2.5

3

3.5

4Pythia/Sherpa

Herwig/Sherpa


MCNLO+Herwig/Sherpa

Figure B.14: Transverse momentum of the fourth leading jet in semileptonic tt events for four common Monte Carlo generators relative topseudo data (Shepra arbitrarily chosen). At the top: reconstructed events with cuts, at the bottom: generated events without cuts. At theleft: before reweighting, at the right: after reweighting.

138 Chapter B. Implementation/Plots of the Reweighting of top pair Events

List of Figures

2.1 Relative strenghts of the four fundamental forces . . . . . . . . . . 62.2 Three leading Feynman diagrams of a top/antitop quark pair pro-

duction by gluon-gluon fusion . . . . . . . . . . . . . . . . . . . . 112.3 Final states of the decay of a top/antitop quark pair . . . . . . . 112.4 Sketch of a proton-proton collision event simulated by Monte Carlo 13

3.1 Schematic view of the LHC accelerator . . . . . . . . . . . . . . . 183.2 Cross section of a LHC dipole magnet . . . . . . . . . . . . . . . . 193.3 Detailed view of the ATLAS detector . . . . . . . . . . . . . . . . 203.4 Schematical event cross section in the ATLAS detector . . . . . . 213.5 Control and data flow of the ATLAS detector . . . . . . . . . . . 223.6 General preconditions for the operation of the ATLAS detector . . 233.7 Calibration and physics analysis in the LCG-GRID . . . . . . . . 243.8 Plan view of a quarter-section of the ATLAS inner detector . . . . 28

4.1 Schematic view of the ATLAS pixel dectector . . . . . . . . . . . 324.2 Schematic view of an ATLAS pixel module . . . . . . . . . . . . . 334.3 Schematic view of the ATLAS pixel dectector readout chain . . . 344.4 Back of Crate card in pixel assembly (Layer-2) . . . . . . . . . . . 364.5 Schematic view of a Back of Crate card . . . . . . . . . . . . . . . 384.6 TX-Plugin, RX-Plugin . . . . . . . . . . . . . . . . . . . . . . . . 404.7 Schematic view of the optical connection between on- and off-

detector side for the ATLAS pixel detector . . . . . . . . . . . . . 424.8 Optical Time Domain Reflectometer . . . . . . . . . . . . . . . . 454.9 Test setup for the optical cables of the ATLAS pixel detector . . . 454.10 OTDR measurement of an optical ribbon from OTDR to ribbon end 464.11 OTDR measurement of an optical ribbon at ribbon end . . . . . . 464.12 Back side of an ATLAS pixel detector readout rack in USA15 be-

fore and after installation of the optical ribbons . . . . . . . . . . 48

5.1 The five track parameters . . . . . . . . . . . . . . . . . . . . . . 545.2 Jet reconstruction flow . . . . . . . . . . . . . . . . . . . . . . . . 60

6.1 Example of a positive jet tag . . . . . . . . . . . . . . . . . . . . . 62

140 LIST OF FIGURES

6.2 Schematic view to illustrate the definition of the signed track im-pact parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6.3 Signed transverse impact parameter distribution for b- and light jets 63

6.4 Dependence of the signed track impact parameter on the jet axis . 64

6.5 Dependence of the signed track impact parameter on the transversemomentum of the track . . . . . . . . . . . . . . . . . . . . . . . . 65

6.6 Dependence of the signed track impact parameter on the pseudo-rapidity of the track . . . . . . . . . . . . . . . . . . . . . . . . . 66

6.7 Illustration of the track impact parameter mapping approach . . . 67

6.8 Signed track impact parameter distributions of Monte Carlo andpseudo data samples . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.9 Mapping of a gaussian distribution on another . . . . . . . . . . . 70

6.10 Signed track impact parameters for tracks with 0.1 < η < 1.9 and1 GeV < pT < 2 GeV . . . . . . . . . . . . . . . . . . . . . . . . . 72

6.11 Track impact parameter correction approach of Borisov and Mar-iotti at DELPHI . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

7.1 The full chain of Monte Carlo simulation in ATLAS . . . . . . . . 78

7.2 Resolution of the transverse momentum of the hadronically decay-ing top quark for Pythia in the range of 200 to 220 GeV . . . . . . 82

7.3 Resolution of the transverse momentum and pseudorapidity of thehadronically decaying top quark for Pythia . . . . . . . . . . . . . 83

7.4 Distributions of the pseudorapidity of the hadronically decayingtop quark on generator level without any cuts for the five chosenMonte Carlo generators . . . . . . . . . . . . . . . . . . . . . . . . 84

7.5 Distributions of the transverse momentum of the hadronically de-caying top quark for five Monte Carlo generators . . . . . . . . . . 87

7.6 The straight line fits to the distribution of the transverse momen-tum of the hadronically decaying top quark for Herwig relative topseudo data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

7.7 Reweighted distributions of the transverse momentum of the ha-dronically decaying top quark for four Monte Carlo generators rel-ative to pseudo data . . . . . . . . . . . . . . . . . . . . . . . . . 90

7.8 The straight line fits to the distribution of the transverse energyHT for Herwig relative to pseudo data . . . . . . . . . . . . . . . 92

7.9 Distributions of the transverse energy HT for four Monte Carlogenerators relative to pseudo data . . . . . . . . . . . . . . . . . . 93

7.10 Chi-square in dependence of the total selection efficiency for fiveMC generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

A.1 Implementation of the signed track impact parameter mappingapproach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

LIST OF FIGURES 141

A.2 Signed transverse impact parameters for tracks with −0.1 < η <0.1, 1 GeV < pT < 2 GeV and 2 GeV < pT < 3 GeV . . . . . . . . 102

A.3 Signed transverse impact parameters for tracks with −0.1 < η <0.1 and 3 GeV < pT < 4 GeV . . . . . . . . . . . . . . . . . . . . . 103

A.4 Signed transverse impact parameters for tracks with 1 GeV < pT <2 GeV, −1.9 < η < −0.1 and 0.1 < η < 1.9 . . . . . . . . . . . . . 104



A.7 Signed transverse impact parameters for tracks with 4 GeV < pT ,−1.9 < η < −0.1 and 0.1 < η < 1.9 . . . . . . . . . . . . . . . . . 107




A.11 Signed transverse impact parameters for tracks with 4 GeV < pT ,−2.5 < η < −1.9 and 1.9 < η < 2.5 . . . . . . . . . . . . . . . . . 111

A.12 Signed longitudinal impact parameters for tracks with −0.1 < η <0.1, 1 GeV < pT < 2 GeV and 2 GeV < pT < 3 GeV . . . . . . . . 112

A.13 Signed longitudinal impact parameters for tracks with −0.1 < η <0.1 and 3 GeV < pT < 4 GeV . . . . . . . . . . . . . . . . . . . . . 113

A.14 Signed longitudinal impact parameters for tracks with 1 GeV <pT < 2 GeV, −1.9 < η < −0.1 and 0.1 < η < 1.9 . . . . . . . . . . 114



A.17 Signed longitudinal impact parameters for tracks with 4 GeV < pT ,−1.9 < η < −0.1 and 0.1 < η < 1.9 . . . . . . . . . . . . . . . . . 117




A.21 Signed longitudinal impact parameters for tracks with 4 GeV < pT ,−2.5 < η < −1.9 and 1.9 < η < 2.5 . . . . . . . . . . . . . . . . . 121

142 LIST OF FIGURES

B.1 Fit functions for the ratio MC/pseudo data of the transverse mo-mentum of the hadronically decaying top quark in semileptonic ttevents for four common Monte Carlo generators . . . . . . . . . . 124

B.2 Transverse momentum of the hadronically decaying top quark insemileptonic tt events for four common Monte Carlo generatorsrelative to pseudo data (Shepra arbitrarily chosen) . . . . . . . . . 125

B.3 Pseudorapidity of the hadronically decaying top quark in semilep-tonic tt events for four common Monte Carlo generators relativeto pseudo data (Shepra arbitrarily chosen) . . . . . . . . . . . . . 126

B.4 Transverse momentum of electrons in semileptonic tt events forfour common Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen) . . . . . . . . . . . . . . . . . . . . . . . . . . 127

B.5 Pseudorapidity of electrons in semileptonic tt events for four com-mon Monte Carlo generators relative to pseudo data (Shepra ar-bitrarily chosen) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

B.6 Transverse momentum of muons in semileptonic tt events for fourcommon Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen) . . . . . . . . . . . . . . . . . . . . . . . . . . 129

B.7 Pseudrapidity of muons in semileptonic tt events for four commonMonte Carlo generators relative to pseudo data (Shepra arbitrarilychosen) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

B.8 Missing transverse energy in semileptonic tt events for four com-mon Monte Carlo generators relative to pseudo data (Shepra ar-bitrarily chosen) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

B.9 Pseudrapidity of jets in semileptonic tt events for four commonMonte Carlo generators relative to pseudo data (Shepra arbitrarilychosen) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

B.10 Transverse momentum of jets in semileptonic tt events for fourcommon Monte Carlo generators relative to pseudo data (Shepraarbitrarily chosen) . . . . . . . . . . . . . . . . . . . . . . . . . . 133

B.11 Transverse momentum of the first leading jet in semileptonic ttevents for four common Monte Carlo generators relative to pseudodata (Shepra arbitrarily chosen) . . . . . . . . . . . . . . . . . . . 134

B.12 Transverse momentum of the second leading jet in semileptonic ttevents for four common Monte Carlo generators relative to pseudodata (Shepra arbitrarily chosen) . . . . . . . . . . . . . . . . . . . 135

B.13 Transverse momentum of the third leading jet in semileptonic ttevents for four common Monte Carlo generators relative to pseudodata (Shepra arbitrarily chosen) . . . . . . . . . . . . . . . . . . . 136

B.14 Transverse momentum of the fourth leading jet in semileptonic ttevents for four common Monte Carlo generators relative to pseudodata (Shepra arbitrarily chosen) . . . . . . . . . . . . . . . . . . . 137

List of Tables

2.1 The three fundamental forces of the Standard Model and theirforce carrier particles. . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 The elementary particles known from the Standard Model . . . . 82.3 The branching ratios of the decay of a top-antitop quark pair . . . 122.4 Basic properties of five common Monte Carlo generators . . . . . 14

3.1 Key numbers of the ATLAS inner detector . . . . . . . . . . . . . 29

4.1 Dimensions and coordinates of the ATLAS pixel dectector subparts 324.2 Back of Crate card types for the ATLAS pixel detector . . . . . . 374.3 Numbers of components for a Back of Crate card . . . . . . . . . 394.4 Current ranges for the Back of Crate card voltage lines . . . . . . 394.5 Status of the Back of Crate card production . . . . . . . . . . . . 404.6 Status of the Back of Crate card installation . . . . . . . . . . . . 414.7 Status of the TX/RX-plugins for the Back of Crate cards . . . . . 424.8 Numbers of optical ribbons for the ATLAS pixel detector . . . . . 444.9 Numbers of optical cables for the ATLAS pixel detector . . . . . . 44

5.1 Resolutions of track parameters in the inner detector . . . . . . . 55

7.1 Configuration of 5 Monte Carlo tt simulations . . . . . . . . . . . 807.2 Total selection efficiencies of semileptonic tt-events . . . . . . . . . 867.3 Total selection efficiencies of semileptonic tt events before and after

reweighting from the transverse momentum of the hadronicallydecaying top quark for five Monte Carlo generators . . . . . . . . 92

7.4 Total selection efficiencies of semileptonic tt events before and afterreweighting from HT for five Monte Carlo generators . . . . . . . 94

144 LIST OF TABLES

Glossary

Notation DescriptionALICE A Lhc Ion Collider Experiment 3, 17AOD Analysis Object Data 23, 51, 79, 99, 123ASIC Application Specific Integrated Circuit 33ATLAS A Toroidal Lhc ApparatuS experiment 3, 17

BE Back-End 23BIS Beam Interlock System 22BOC Back Of Crate card 34BPM Bi-Phase Mark 35

CAN Controller Area Network 23CERN European Organization for Nuclear Research 3, 17CKM Cabibbo-Kobayashi-Maskawa 9CMS Compact Muon Solenoid experiment 3, 17CPLD Complex Programmable Logic Device 37, 38CSC Cathode Strip Chamber 25

DAQ Data AcQuisition 22DCS Detector Control System 22DPD Derived Physics Data 23, 51, 79DSS Detector Safety System 22

EB Event Builder 23ECAL Electromagnetic CALorimeter 26EF Event Filter 23ELMB Embedded Local Monitor Board 23ESD ElectroStatic Discharge 41ESD Event Summary Data 23, 51, 79

FCAL Forward CALorimeter 26FE Front-End chip of pixel detector 33FE Front-End of DCS 23

146 Glossary

Notation DescriptionFPIAA Finding Persons Inside Atlas Area 22

GCS Global Control Station 23GEANT4 GEometry ANd Tracking version 4 77GRIN GRadient INdex 42, 48

HCAL Hadronic CALorimeter 26HEC Hadronic Endcap Calorimeter 27HLT High-Level Trigger 23HOLA High speed Optical Link for Atlas 37HV High Voltage 33

ID Inner Detector 27IP Impact Parameter 61

L1A Level-1 Accept 23LAr Liquid Argon 27LCG Lhc Computing Grid 23LCS Local Control Stations 23LHC Large Hadron Collider 3, 9, 17LHCb LHC beauty experiment 3, 17LLA Leading Logarithmic Accuracy 13LO Leading Order 12LVL1 LeVeL-1 trigger 22LVL2 LeVeL-2 trigger 23

MC Monte Carlo 4, 5, 51MCC Module Control Chip 33MDT Monitored Drift-Tube chamber 25MM Multi Mode 47MT Mechanical Transfer 43

NLO Next to Leading Order 12NTC Negative Temperature Coefficient 33

OB OptoBoard 33, 34OTDR Optical Time Domain Reflectometer 44, 47

PC Personal Computer 44PDF Parton Distribution Function 12PIN P-type, Intrinsic, N-type 37PP Patch Panel 39

Glossary 147

Notation DescriptionPS Proton Synchrotron 17PV Primary Vertex 61

QCD Quantum Chromo-Dynamics 7QED Quantum Electro-Dynamics 8

RCC Readout Crate Controller 34RD Real Data 66RDO Raw Data Object 23, 51, 79ROB Read Out Buffer 23, 35ROD Read Out Driver 23, 34RoI Regions of Interest 23RoIB Regions of Interest Builder 23ROOT “ROOT is an object-oriented framework

aimed at solving the data analysis challengesof high-energy physics”

99, 123

RPC Resistive Plate Chamber 25RX Receiver 37, 39

SBC Single Board Computer 34SCS Sub-detector Control Station 23SCT SemiConductor Tracker 27, 35SIMM Stepped Index Multi Mode 42, 48SM Single Mode 44SM Standard Model 5SMC Sub Miniature C 43SPS Super Proton Synchrotron 17ST Stab and Twist 44SV Secondary Vertex 61

TDAQ Trigger and Data AcQuisition 22TGC Thin Gap Chamber 25TIM Ttc Interface Module 34TMT Thermal Management Tile 33TRT Transition Radiation Tracker 27TTC Timing, Trigger and Control 22, 34TX Transmitter 37, 39

USB Universal Serial Bus 44

VCSEL Vertical Cavity Surface Emitting Laser 37VME VersaModule Eurocard 34

148 Glossary

Bibliography

[1] R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures onPhysics, Addison-Wesley 1998.

[2] R. P. Feynman, The Meaning of It All, Perseus Books 1998.

[3] J. Rohlf, Modern Physics from α to Z0, John Wiley & Sons 1994.

[4] ALEPH, DELPHI, L3 and OPAL Collaborations, Phys. Lett. B565 (2003)61.

[5] Combined CDF and DZero Upper Limits on Standard Model Higgs-BosonProduction with up to 4.2 fb-1 of Data, FERMILAB-PUB-09-060-E 2009http://arxiv.org/abs/0903.4001

[6] Design report Tevatron 1 Project. FERMILAB-DESIGN-1982-01, 1982http://lss.fnal.gov/archive/design/fermilab-design-1982-01.

pdf

http://www-bdnew.fnal.gov/tevatron/

[7] CDF Collaboration, Phys. Rev. D50 (1994) 2966 and Phys. Rev. D52(1995) 4784.http://www-cdf.fnal.gov/

[8] DØ Collaboration, Phys. Rev. Lett. 77 (1996) 3309.http://www-d0.fnal.gov/

[9] Combination of CDF and DØ Results on the Mass of the Top Quark,FERMILAB-TM-2427-E 2009http://arxiv.org/abs/0903.2503

http://tevewwg.fnal.gov/top/

[10] R. Bonciani, S. Catani, M. L. Mangano, and P. Nason, “NLL resumma-tion of the heavy-quark hadroproduction cross-section”, Nucl. Phys. B529(1998) 424–450.http://arxiv.org/abs/hep-ph/9801375

http://arxiv.org/abs/0903.4001

http://lss.fnal.gov/archive/design/fermilab-design-1982-01.pdf

http://lss.fnal.gov/archive/design/fermilab-design-1982-01.pdf

http://www-bdnew.fnal.gov/tevatron/

http://www-cdf.fnal.gov/

http://www-d0.fnal.gov/

http://arxiv.org/abs/0903.2503

http://tevewwg.fnal.gov/top/

http://arxiv.org/abs/hep-ph/9801375

150 BIBLIOGRAPHY

[11] T. Flick, Studies on the Optical Readout for the ATLAS Pixel Detector,PhD thesis, Universitat Wuppertal, 2006.

[12] Particle Data Group, The European Physical Journal B 667,1 (2008)http://pdg.lbl.gov

[13] E. Thomson, Top Quark Pair Production Cross Section Combinationhttp://www.physics.upenn.edu/~thomsone/2006_1030_dpf_v5-1.

pdf

[14] Dobbs, Frixione, Laenen, Tollefson: Les Houches Guidebook to MonteCarlo Generators for Hadron Collider Physics. 2004http://arxiv.org/abs/hep-ph/0403045

[15] F. Krauss: Sketch of a proton-proton collision event simulated by MonteCarlo.

[16] Parton Fragmentation and String Dynamics, Bo Andersson, G. Gustafson,G. Ingelman, T. Sjostrand (Lund U., Dept. Theor. Phys.), LU-TP-83-10,Jul 1983. Phys.Rept.97:31-145,1983.

[17] G. Corcella, I. G. Knowles, G. Marchesini, S. Moretti, K. Odagiri, P.Richardson, M. H. Seymour, B. R. Webber: HERWIG 6.5: an event gen-erator for Hadron Emission Reactions With Interfering Gluons (includingsupersymmetric processes), JHEP 0101 (2001) 010.http://arxiv.org/abs/hep-ph/0011363

http://hepwww.rl.ac.uk/theory/seymour/herwig/

[18] T. Sjostrand, S. Mrenna, P. Skands, Pythia 6.4 Physics and Manual, LUTP 06-13, FERMILAB-PUB-06-052-CD-T, 2006.http://home.thep.lu.se/~torbjorn/Pythia.html

[19] B. Kersevan, E. Richter-Was: The Monte Carlo Event Generator AcerMC2.0 with Interfaces to PYTHIA 6.2 and HERWIG 6.5, Comput. Phys.Commun. 149 (2003) 142.http://arxiv.org/abs/hep-ph/0405247

http://borut.home.cern.ch/borut/

[20] T. Gleisberg et al.: Event generation with SHERPA 1.1, (2008), SLAC-PUB-13420http://arxiv.org/pdf/0811.4622

http://www.sherpa-mc.de/

[21] S. Frixione and B. R. Webber, Matching NLO QCD computations andparton shower simulations, JHEP 0206 (2002) 029.http://www.hep.phy.cam.ac.uk/theory/webber/MCatNLO/

http://pdg.lbl.gov

http://www.physics.upenn.edu/~thomsone/2006_1030_dpf_v5-1.pdf

http://www.physics.upenn.edu/~thomsone/2006_1030_dpf_v5-1.pdf



http://hepwww.rl.ac.uk/theory/seymour/herwig/

http://home.thep.lu.se/~torbjorn/Pythia.html


http://borut.home.cern.ch/borut/

http://arxiv.org/pdf/0811.4622

http://www.sherpa-mc.de/

http://www.hep.phy.cam.ac.uk/theory/webber/MCatNLO/

BIBLIOGRAPHY 151

[22] UA5 Collaboration, G. J. Alner et al., Nucl. Phys. B 291, (1987) 445.

[23] J. M. Butterworth, J. R. Forshaw, M. H. Seymour, Z. Phys. C 72 (1996)637.http://arxiv.org/abs/hep-ph/9601371

[24] LHC Design Report, CERN-2004-003-V-1http://cdsweb.cern.ch/record/782076

http://ab-div.web.cern.ch/ab-div/Publications/

LHC-DesignReport.html

http://lhc.web.cern.ch/lhc/

[25] ALICE physics performance: Technical Design Report, ALICE-TDR-13,CERN-LHCC-2005-030http://cdsweb.cern.ch/record/879894

http://aliceinfo.cern.ch

[26] ATLAS detector and physics performance: Technical Design Report1+2, ATLAS-TDR-014, CERN-LHCC-99-014; ATLAS-TDR-015, CERN-LHCC-99-015http://cdsweb.cern.ch/record/391176

http://cdsweb.cern.ch/record/391177

http://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/TDR/access.

html

http://www.atlas.ch/

[27] CMS physics: Technical Design Report 1+2, CERN-LHCC-2006-001,CMS-TDR-008-1; CERN-LHCC-2006-021, CMS-TDR-008-2http://cdsweb.cern.ch/record/922757


http://cms.cern.ch

[28] LHCb reoptimized detector design and performance: Technical Design Re-port, CERN-LHCC-2003-030, LHCb-TDR-9http://cdsweb.cern.ch/record/630827

http://lhcb.web.cern.ch/

[29] Proton Synchrotron at CERN webpage:http://www.cern.ch/CERN/Divisions/PS/Welcome.html

[30] Super Synchrotron at CERN webpage:http://ab-dep-op-sps.web.cern.ch/ab-dep-op-sps/

[31] Overall view of LHC experimentshttp://cdsweb.cern.ch/record/841555?ln=en



http://ab-div.web.cern.ch/ab-div/Publications/LHC-DesignReport.html

http://ab-div.web.cern.ch/ab-div/Publications/LHC-DesignReport.html

http://lhc.web.cern.ch/lhc/


http://aliceinfo.cern.ch



http://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/TDR/access.html

http://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/TDR/access.html

http://www.atlas.ch/



http://cms.cern.ch


http://lhcb.web.cern.ch/

http://www.cern.ch/CERN/Divisions/PS/Welcome.html

http://ab-dep-op-sps.web.cern.ch/ab-dep-op-sps/

http://cdsweb.cern.ch/record/841555?ln=en

152 BIBLIOGRAPHY

[32] Cross section of the vacuum pipes inside a dipole magnethttps://edms.cern.ch/document/101865/1.1

[33] LHC luminosity webpage:http://ab-dep-bi-pm.web.cern.ch/ab-dep-bi-pm/?n=Activities.

BRAN

[34] ATLAS Collaboration, G. Aad et al., The ATLAS Experiment at theCERN Large Hadron Collider, JINST 3 (2008) S08003.https://twiki.cern.ch/twiki/pub/Atlas/AtlasTechnicalPaper/

Published_version_jinst8_08_s08003.pdf

https://twiki.cern.ch/twiki/bin/view/Atlas/

AtlasTechnicalPaperListOfFigures

[35] Computer generated image of the whole ATLAS detectorhttp://mediaarchive.cern.ch/MediaArchive/Photo/Public/2008/

0803012/0803012_01/0803012_01-Icon.jpg

[36] ATLAS inner detector: Technical Design Report 1+2, ATLAS-TDR-004,CERN-LHCC-97-016; ATLAS-TDR-005, CERN-LHCC-97-017http://cdsweb.cern.ch/record/331063


https://twiki.cern.ch/twiki/bin/view/Atlas/InnerDetector

[37] ATLAS calorimeter performance: Technical Design Report, ATLAS-TDR-001, CERN-LHCC-96-040http://cdsweb.cern.ch/record/331059

http://atlas.web.cern.ch/Atlas/TDR/caloperf/caloperf.html

[38] ATLAS muon spectrometer: Technical Design Report, ATLAS-TDR-010,CERN-LHCC-97-022http://cdsweb.cern.ch/record/331068

http://atlas.web.cern.ch/Atlas/GROUPS/MUON/TDR/Web/TDR.html

[39] Event Cross Section in a computer generated image of the ATLAS detector,CERN-GE-0803022http://cdsweb.cern.ch/record/1096081

[40] ATLAS magnet system: Technical Design Report, ATLAS-TDR-006,CERN-LHCC-97-018http://cdsweb.cern.ch/record/338080

[41] ATLAS central solenoid: Technical Design Report, ATLAS-TDR-009,CERN-LHCC-97-021http://cdsweb.cern.ch/record/331067

https://edms.cern.ch/document/101865/1.1

http://ab-dep-bi-pm.web.cern.ch/ab-dep-bi-pm/?n=Activities.BRAN

http://ab-dep-bi-pm.web.cern.ch/ab-dep-bi-pm/?n=Activities.BRAN

https://twiki.cern.ch/twiki/pub/Atlas/AtlasTechnicalPaper/Published_version_jinst8_08_s08003.pdf

https://twiki.cern.ch/twiki/pub/Atlas/AtlasTechnicalPaper/Published_version_jinst8_08_s08003.pdf

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasTechnicalPaperListOfFigures

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasTechnicalPaperListOfFigures

http://mediaarchive.cern.ch/MediaArchive/Photo/Public/2008/0803012/0803012_01/0803012_01-Icon.jpg

http://mediaarchive.cern.ch/MediaArchive/Photo/Public/2008/0803012/0803012_01/0803012_01-Icon.jpg



https://twiki.cern.ch/twiki/bin/view/Atlas/InnerDetector


http://atlas.web.cern.ch/Atlas/TDR/caloperf/caloperf.html


http://atlas.web.cern.ch/Atlas/GROUPS/MUON/TDR/Web/TDR.html




BIBLIOGRAPHY 153

[42] ATLAS barrel toroid: Technical Design Report, ATLAS-TDR-007, CERN-LHCC-97-019http://cdsweb.cern.ch/record/331065

[43] ATLAS endcap toroids: Technical Design Report, ATLAS-TDR-008,CERN-LHCC-97-020http://cdsweb.cern.ch/record/331066

[44] S. Baranov, M. Bosman, I. Dawson, V. Hedberg, A. Nisati, M. Shupe,Estimation of Radiation Background, Impact on Detectors, Activation andShielding Optimization in ATLAS, ATL-GEN-2005-001, ATL-COM-GEN-2005-001, CERN-ATL-GEN-2005-001http://cdsweb.cern.ch/record/814823

[45] Hierarchical Control of the ATLAS Experiment, CERN-THESIS-2007-037http://cdsweb.cern.ch/record/1034400

[46] Burckhart, H. J., Cook, J., Filiminov, V., Franz, T., Gutzwiller, O., Kho-mutnikov, V., Schlenker, S., The Common Infrastructure Control of theATLAS experiment, Topical Workshop on Electronics for Particle Physics,Naxos, Greece, 15–19 Sep 2008, pp.428–431http://cdsweb.cern.ch/record/1159528

[47] Machine Protection for the LHC: Architecture of the Beam and PoweringInterlock Systems, LHC-Project-Report-521, CERN-LHC-Project-Report-521http://cdsweb.cern.ch/record/531820

[48] B.G. Taylor: Timing distribution at the LHC. Workshop on Electronicsfor LHC Experiments, 2002.http://cdsweb.cern.ch/record/592719

[49] The ATLAS central level-1 trigger logic and TTC system, ATL-COM-DAQ-2008-006http://cdsweb.cern.ch/record/1119310

[50] C. Cardeira, O. Beltramello, and H. J. Burckhart et al., CommunicationArchitecture of a system to find persons inside ATLAS, proceedings of the6th IEEE International Workshop on Factory Communication Systems,Torino (2006) 405–409

[51] O. Beltramello, H. Burckhart, S. Franz, M. Jaekel, M. Jeckel, S. Lued-ers, G. Morpurgo, F. dos Santos Pedrosa, K. Pommes, H. Sandaker, TheATLAS Detector Safety System









154 BIBLIOGRAPHY

[52] A. Barriuso Poy et al., The detector control system of the ATLASexperiment, JINST 3 (2008) P05006http://www.iop.org/EJ/article/1748-0221/3/05/P05006/jinst8_

05_p05006.pdf

http://atlas-dcs.web.cern.ch/atlas-dcs/

[53] ATLAS high-level trigger, data-acquisition and controls: Technical DesignReport, ATLAS-TDR-016, CERN-LHCC-2003-022http://cdsweb.cern.ch/record/616089

[54] ATLAS DAQ and Controls, ATL-DAQ-SLIDE-2009-094, ATL-COM-DAQ-2009-032http://cdsweb.cern.ch/record/1174769

[55] ATLAS level-1 trigger: Technical Design Report, ATLAS-TDR-012,CERN-LHCC-98-014http://cdsweb.cern.ch/record/381429

[56] ATLAS computing: Technical Design Report, ATLAS-TDR-017, CERN-LHCC-2005-022http://cdsweb.cern.ch/record/837738

http://atlas-proj-computing-tdr.web.cern.ch/

atlas-proj-computing-tdr/Html/

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasComputing

[57] The ATLAS Analysis Architecture, ATL-SLIDE-2007-028, CERN-ATL-SLIDE-2007-028http://cdsweb.cern.ch/record/1047631

[58] LHC computing Grid: Technical Design Report, CERN-LHCC-2005-024,LCG-TDR-001http://cdsweb.cern.ch/record/840543

http://lcg.web.cern.ch/LCG/

[59] J. Shiers, The Worldwide LHC Computing Grid (LCG), Comput. Phys.Commun. 177, 1–2 (2007) 219–223.http://cdsweb.cern.ch/record/1063847

[60] The ATLAS Computing Model, ATL-SOFT-2004-007, ATL-COM-SOFT-2004-009, CERN-ATL-COM-SOFT-2004-009, CERN-LHCC-2004-037-G-085http://cdsweb.cern.ch/record/811058

[61] The C++ Standards Committee webpage:http://www.open-std.org/jtc1/sc22/wg21/

http://www.iop.org/EJ/article/1748-0221/3/05/P05006/jinst8_05_p05006.pdf

http://www.iop.org/EJ/article/1748-0221/3/05/P05006/jinst8_05_p05006.pdf

http://atlas-dcs.web.cern.ch/atlas-dcs/





http://atlas-proj-computing-tdr.web.cern.ch/atlas-proj-computing-tdr/Html/

http://atlas-proj-computing-tdr.web.cern.ch/atlas-proj-computing-tdr/Html/

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasComputing



http://lcg.web.cern.ch/LCG/



http://www.open-std.org/jtc1/sc22/wg21/

BIBLIOGRAPHY 155

[62] B. Stroustrup, The C++ Programming Language. Addison-Wesley 2000.

[63] The Fortran Standards webpage:http://www.nag.co.uk/sc22wg5/

[64] The Java Community Process Programhttp://www.jcp.org/en/home/index

[65] Python Programming Language webpage:http://www.python.org/

[66] ATLAS Athena Framework wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/AthenaFramework

[67] ATLAS tile calorimeter: Technical Design Report, ATLAS-TDR-003,CERN-LHCC-96-042http://cdsweb.cern.ch/record/331062

[68] ATLAS liquid-argon calorimeter: Technical Design Report, ATLAS-TDR-002, CERN-LHCC-96-041http://cdsweb.cern.ch/record/331061

[69] M. Capeans-Garrido, The transition radiation tracker of the ATLAS ex-periment, IEEE Trans. Nucl. Sci. 51 (2004) 994–1000.http://cdsweb.cern.ch/record/816797

[70] J. N. Jackson, The ATLAS semiconductor tracker (SCT), Nucl. Instrum.Methods Phys. Res., A 541 (2005) 89–95.http://cdsweb.cern.ch/record/908854?ln=en

ATLAS SCT detector webpage:http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/SCT/

ATLAS SCT detector wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/SctWiki

[71] ATLAS pixel detector: Technical Design Report, ATLAS-TDR-011,CERN-LHCC-98-013http://cdsweb.cern.ch/record/381263

ATLAS Pixel detector webpage:http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/PIXELS/

pixel.html

ATLAS Pixel wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/PixelWiki

[72] A 3D model of the pixel detector and it’s frameworkhttp://hep.phys.sfu.ca/openhouse_2008/kiosk/gallery/atlas_

photos/selected-photos/inner-detector/pixel/Pixels_Complete_

Opened.jpg

http://www.nag.co.uk/sc22wg5/

http://www.jcp.org/en/home/index

http://www.python.org/

https://twiki.cern.ch/twiki/bin/view/Atlas/AthenaFramework




http://cdsweb.cern.ch/record/908854?ln=en

http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/SCT/

https://twiki.cern.ch/twiki/bin/view/Atlas/SctWiki


http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/PIXELS/pixel.html

http://atlas.web.cern.ch/Atlas/GROUPS/INNER_DETECTOR/PIXELS/pixel.html

https://twiki.cern.ch/twiki/bin/view/Atlas/PixelWiki

http://hep.phys.sfu.ca/openhouse_2008/kiosk/gallery/atlas_photos/selected-photos/inner-detector/pixel/Pixels_Complete_Opened.jpg



156 BIBLIOGRAPHY

[73] F. Hugging et al., Design and test of pixel sensors for operation in severeradiation environments, Nucl. Instrum. Meth., A439 (2000) 529.

[74] T. Stockmanns, Multi-Chip-Modul-Entwicklung fur den ATLAS Pixelde-tektor: Analyse der Front-End-Chip-Elektronik in strahlenharter 0,25-µm-Technologie und Entwicklung und Realisierung eines Serial-Powering-Konzeptes, PhD thesis, Universitat Bonn, July 2004.

[75] E. Mandelli et al., Digital Column Readout Architecture for the ATLASPixel 0.25 µm Front End IC, IEEE, 0-7803-7324-3/02 2002.http://ieeexplore.ieee.org/iel5/7884/21725/01008527.pdf

[76] R. Beccherle et al., The Module Controller Chip for the ATLAS PixelDetector, NIM-A, 492 (2002) 117–133.

[77] K. E. Arms et al., ATLAS pixel opto-electronics, NIM-A, 554 (2005) 458–468.http://www.physics.ohio-state.edu/~gan/optoboard/NIM.pdf

[78] BOC card documentation webpage:http://www.atlas.uni-wuppertal.de/~bocprod/boc_docu.html

[79] ROD card documentation webpage:http://wisconsin.cern.ch/ROD/ATLAS_SiRod.shtml

[80] TIM card documentation webpage:http://www.hep.ucl.ac.uk/atlas/sct/tim/

[81] I.-M. Gregor, Optical Links for the ATLAS pixel detector, PhD thesis,Universitat Wuppertal, 2001.

[82] Cavendish Laboratory Cambridge webpage:http://www.phy.cam.ac.uk/

[83] M. L. Chu et al., The off-detector opto-electronics for the optical linksof the ATLAS semiconductor tracker and pixel detector, Nucl. Instrum.Meth., A530 (2004) 293–310.http://dialnet.unirioja.es/servlet/articulo?codigo=972163

[84] E. van der Bij, R. A. McLaren, O. Boyle and G. Rubin: S-LINK, a datalink interface specification for the LHC era, IEEE Trans. Nucl. Sci., 44(1997) 398–402.

[85] The CERN S-Link webpage:http://hsi.web.cern.ch/HSI/s-link/

http://ieeexplore.ieee.org/iel5/7884/21725/01008527.pdf

http://www.physics.ohio-state.edu/~gan/optoboard/NIM.pdf

http://www.atlas.uni-wuppertal.de/~bocprod/boc_docu.html

http://wisconsin.cern.ch/ROD/ATLAS_SiRod.shtml

http://www.hep.ucl.ac.uk/atlas/sct/tim/

http://www.phy.cam.ac.uk/

http://dialnet.unirioja.es/servlet/articulo?codigo=972163

http://hsi.web.cern.ch/HSI/s-link/

BIBLIOGRAPHY 157

[86] Turck Elektronik/Fertigung webpage:http://electronic-production.eu/

[87] M. Goodrick: BOC - The Back of Crate interface for the SCT ROD.Technical report, Cavendish-Laboratory, Cambridge, 2003.

[88] BOC card production webpage:http://www.atlas.uni-wuppertal.de/~bocprod/boc_production.

php

[89] Pixel Production Database:http://www.ge.infn.it/~atlas/pdb/

[90] ATLAS pixel connectivity table:https://edms.cern.ch/file/447936/3/ATL-IP-ES-0102_AppendixA_

v3.91.xls

[91] BOC/ROD installation tables:https://twiki.cern.ch/twiki/pub/Atlas/

CountingRoomsInstallation/AtlasPixelUSA15BOCRODInstallationVer1.

xls

[92] Andreas Korn’s talk about DAQ crate signoff:http://indico.cern.ch/getFile.py/access?contribId=2&resId=0&

materialId=slides&confId=23435

[93] ATLAS pixel TX-plugin failures wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/PixelTxPlugins

[94] G. Lenzen, Pixel Opto Services, ATL-IP-ES-0070, January 2004.

[95] Arvidsson, Bjork, Pearce, Troska, Vasey, Zanet: A Dense Multi-RibbonCable For Installation in a Harsh Environment at CERN, 2000.

[96] Luciol ν-OTDRhttp://www.luciol.com/

[97] Wiki of ATLAS pixel connectivity test in SR1-building:https://twiki.cern.ch/twiki/bin/view/Atlas/

ConnectivityTestResults

[98] ATLAS pixel connectivity table of optical cables:https://twiki.cern.ch/twiki/pub/Atlas/PixelConnectivity/

ATL-IP-ES-0102_ver4.1_Stephan_Georg20071024.xls

https://twiki.cern.ch/twiki/bin/viewfile/Atlas/

PixelConnection?rev=1;filename=ATL-IP-ES-0102_ver4.1_

Stephan_Georg20071024_swaps_V1_CorrectDB_AfterTestFinal.

xls

http://electronic-production.eu/

http://www.atlas.uni-wuppertal.de/~bocprod/boc_production.php

http://www.atlas.uni-wuppertal.de/~bocprod/boc_production.php

http://www.ge.infn.it/~atlas/pdb/

https://edms.cern.ch/file/447936/3/ATL-IP-ES-0102_AppendixA_v3.91.xls

https://edms.cern.ch/file/447936/3/ATL-IP-ES-0102_AppendixA_v3.91.xls

https://twiki.cern.ch/twiki/pub/Atlas/CountingRoomsInstallation/AtlasPixelUSA15BOCRODInstallationVer1.xls



http://indico.cern.ch/getFile.py/access?contribId=2&resId=0&materialId=slides&confId=23435


https://twiki.cern.ch/twiki/bin/view/Atlas/PixelTxPlugins

http://www.luciol.com/

https://twiki.cern.ch/twiki/bin/view/Atlas/ConnectivityTestResults

https://twiki.cern.ch/twiki/bin/view/Atlas/ConnectivityTestResults

https://twiki.cern.ch/twiki/pub/Atlas/PixelConnectivity/ATL-IP-ES-0102_ver4.1_Stephan_Georg20071024.xls

https://twiki.cern.ch/twiki/pub/Atlas/PixelConnectivity/ATL-IP-ES-0102_ver4.1_Stephan_Georg20071024.xls

https://twiki.cern.ch/twiki/bin/viewfile/Atlas/PixelConnection?rev=1;filename=ATL-IP-ES-0102_ver4.1_Stephan_Georg20071024_swaps_V1_CorrectDB_AfterTestFinal.xls




158 BIBLIOGRAPHY

[99] Test results of the ATLAS pixel optical cables:https://twiki.cern.ch/twiki/pub/Atlas/PixelConnectivity/

Connectivity.doc

[100] Data of the OTDR test of the optical cables of the ATLAS pixel detector:/castor/cern.ch/user/s/sandvoss/atlas/hardware/pixel/

OptoCablesRibbonsFibres/pit/USA15/afterribboninstall

[101] ATLAS pixel connectivity test wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/

PixelConnectivityTest

[102] Optical fiber connection test results:http://indico.cern.ch/getFile.py/access?contribId=0&resId=0&


[103] ATLAS Reconstruction wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/

ReconstructionDocumentation

[104] T. Cornelissen et al., Concepts, Design and Implementation of the ATLASNew Tracking, ATLAS Note ATL-SOFT-PUB-2007-007 (2007).http://cdsweb.cern.ch/record/1020106

[105] V. Kartvelishvili, Electron bremsstrahlung recovery in ATLAS, Nucl. Phys.172 (Proc. Suppl.) (2007) 208.

[106] R. Fruhwirth, Track fitting with non-Gaussian noise, Comput. Phys.Comm. 100 (1997) 1.

[107] R. Fruhwirth and A. Strandlie, Track fitting with ambituities and noise:a study of elastic tracking and nonlinear filters, Comput. Phys. Commun.120 (1999) 197.

[108] F. Bergsma, Calibration of Hall sensors in three dimensions, in Proceedingsof the 13th international magnetic measurement workshop, Stanford USA(2003).http://cdsweb.cern.ch/record/1072471

[109] M. Aleksa et al., Measurement of the ATLAS solenoid magnetic field, 2007JINST 3 P04003.

[110] A. Salzburger: A Parametrization for Fast Simulation of Muon Tracks inthe ATLAS Inner Detector and Muon System, Diploma Thesis Innsbruck2003http://cdsweb.cern.ch/record/813003

https://twiki.cern.ch/twiki/pub/Atlas/PixelConnectivity/Connectivity.doc

https://twiki.cern.ch/twiki/pub/Atlas/PixelConnectivity/Connectivity.doc

/castor/cern.ch/user/s/sandvoss/atlas/hardware/pixel/OptoCablesRibbonsFibres/pit/USA15/afterribboninstall

/castor/cern.ch/user/s/sandvoss/atlas/hardware/pixel/OptoCablesRibbonsFibres/pit/USA15/afterribboninstall

https://twiki.cern.ch/twiki/bin/view/Atlas/PixelConnectivityTest

https://twiki.cern.ch/twiki/bin/view/Atlas/PixelConnectivityTest



https://twiki.cern.ch/twiki/bin/view/Atlas/ReconstructionDocumentation

https://twiki.cern.ch/twiki/bin/view/Atlas/ReconstructionDocumentation




BIBLIOGRAPHY 159

[111] P. Billoir, S. Quan, Fast vertex fitting with a local parametrization of tracks,Nucl. Instrum. Meth. A311 (1992) 139–150.

[112] R. Fruhwirth, Application of Kalman Filtering to Track and Vertex Fit-ting, Nucl. Inistrum. and Methods 225 (1984) 352.

[113] I. Abt et al., The tracking, calorimeter and muon detectors of the H1experiment at HERA, Nucl. Instrum. Meth. A 386 (1997) 348.

[114] ATLAS Collaboration, Expected Performance of the ATLAS Experiment;Detector, Trigger and Physics, arXiv:0901.0512v3. April 2009.

[115] H. Bachacou: Mistags at CDF. ATLAS Flavour Tagging Workshop 2007.http://indico.cern.ch/getFile.py/access?contribId=64&

sessionId=12&resId=0&materialId=slides&confId=14475

[116] C. Gwilliam, S. Burdin, Track Impact Parameter Tuninghttp://indico.cern.ch/getFile.py/access?contribId=1&resId=0&


[117] MC AOD dataset 5200 (tt simulated with MC@NLO, Jimmy and Herwig)with 6 · 105 events, reconstructed with athena 13.0.30.4 with misaligneddetector and trigger information:trig1_misal1_mc12.005200.T1_McAtNlo_Jimmy.recon.AOD.

v13003004_tid018416

[118] MC AOD dataset 5200 (tt simulated with MC@NLO, Jimmy and Herwig)with 5 · 104 events, reconstructed with athena 13.0.30.4 with heavilymisaligned detector (special V6) and trigger information:special1_misal1_mc12_V6.005200.T1_McAtNlo_Jimmy.recon.AOD.

v13003004

[119] G. Borisov, C. Mariotti, Fine tuning of track impact parameter resolutionof the DELPHI detector, Nucl. Inst. Meth A372 (1996) 181.

[120] Burdin, Anzelc, Decay Length Resolution Studies, Application to B0s Mix-

ing and p17 Calibration. DØ Note 5336, 2007

[121] G. Cowan, Statistical data analysis, Oxford Science Publications 2002.

[122] ATLAS Generation wiki:https://twiki.cern.ch/twiki/bin/view/AtlasProtected/

MonteCarloWorkingGroup

[123] J. Allison et al., Geant4 Developments and Applications, IEEE Transac-tions on Nuclear Science 53 No. 1 (2006) 270–278.S. Agostinelli et al., Geant4 - A Simulation Toolkit, Nuclear Instruments

http://indico.cern.ch/getFile.py/access?contribId=64&sessionId=12&resId=0&materialId=slides&confId=14475




trig1_misal1_mc12.005200.T1_McAtNlo_Jimmy.recon.AOD.v13003004_tid018416

trig1_misal1_mc12.005200.T1_McAtNlo_Jimmy.recon.AOD.v13003004_tid018416

special1_misal1_mc12_V6.005200.T1_McAtNlo_Jimmy.recon.AOD.v13003004

special1_misal1_mc12_V6.005200.T1_McAtNlo_Jimmy.recon.AOD.v13003004

https://twiki.cern.ch/twiki/bin/view/AtlasProtected/MonteCarloWorkingGroup

https://twiki.cern.ch/twiki/bin/view/AtlasProtected/MonteCarloWorkingGroup

160 BIBLIOGRAPHY

and Methods A 506 (2003) 250–303.http://geant4.web.cern.ch/geant4/

[124] ATLFAST 2.0 a fast simulation package for ATLAS, ATL-PHYS-98-131http://cdsweb.cern.ch/record/683751

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlfastII

http://indico.cern.ch/getFile.py/access?contribId=82&

sessionId=22&resId=0&materialId=slides&confId=52623

[125] ATLAS Digitization wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasDigitization

[126] ATLAS Simulation wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasSimulation

[127] ATLFAST 1.0 A package for particle-level analysis, ATL-PHYS-96-079,ATL-GE-PN-79http://cdsweb.cern.ch/record/682460

https://twiki.cern.ch/twiki/bin/view/Atlas/

AtlfastDocumentation

[128] The Fast ATLAS Track Simulation (FATRAS), ATL-SOFT-PUB-2008-001, ATL-COM-SOFT-2008-002http://cdsweb.cern.ch/record/1091969

https://twiki.cern.ch/twiki/bin/view/Atlas/FatRas

[129] J. Pumplin et al., JHEP 07 (2002) 012.

[130] J. Pumplin, A. Belyaev, J. Huston, D. Stump and W. K. Tung, JHEP 02(2006) 032.

[131] The 5 MC top/antitop datasets:grid-ui.physik.uni-wuppertal.de:/sfsscratch/atlas/sandvoss/

data/atlas/software/mcgenerators/

[132] The scripts for the production of the 5 MC top/antitop datasets:grid-ui.physik.uni-wuppertal.de:/sfsscratch/atlas/sandvoss/

data/atlas/software/mcgenerators/

[133] Object selection of the ATLAS top physics group:https://twiki.cern.ch/twiki/bin/view/AtlasProtected/

TopGroupCSCObjectSelection

[134] ATLAS Quality Assurance Group, ATLAS C++ Coding Standard, 2003http://atlas-computing.web.cern.ch/atlas-computing/projects/

qa/AtlasCCS1_2_1.pdf

http://geant4.web.cern.ch/geant4/


https://twiki.cern.ch/twiki/bin/view/Atlas/AtlfastII



https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasDigitization

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasSimulation


https://twiki.cern.ch/twiki/bin/view/Atlas/AtlfastDocumentation

https://twiki.cern.ch/twiki/bin/view/Atlas/AtlfastDocumentation


https://twiki.cern.ch/twiki/bin/view/Atlas/FatRas

grid-ui.physik.uni-wuppertal.de:/sfsscratch/atlas/sandvoss/data/atlas/software/mcgenerators/




https://twiki.cern.ch/twiki/bin/view/AtlasProtected/TopGroupCSCObjectSelection

https://twiki.cern.ch/twiki/bin/view/AtlasProtected/TopGroupCSCObjectSelection

http://atlas-computing.web.cern.ch/atlas-computing/projects/qa/AtlasCCS1_2_1.pdf

http://atlas-computing.web.cern.ch/atlas-computing/projects/qa/AtlasCCS1_2_1.pdf

BIBLIOGRAPHY 161

[135] ATLAS Improving Software wiki:https://twiki.cern.ch/twiki/bin/view/Atlas/ImprovingSoftware

[136] Linux Kernel Coding Style:/usr/src/linux/Documentation/CodingStyle

[137] Kernighan, Pike, The Practice of Programming, Addison-Wesley 1999

[138] The Athena package ImpactParameterTuning:http://atlas-sw.cern.ch/cgi-bin/viewcvs-atlas.cgi/

offline/PhysicsAnalysis/JetTagging/JetTagCorrections/

ImpactParameterTuning

[139] CERN ROOT webpage:http://root.cern.ch/drupal/

[140] The Athena package DifferentialCrossSectionTTbar:http://atlas-sw.cern.ch/cgi-bin/viewcvs-atlas.cgi/users/

sandvoss/DifferentialCrossSectionTTbar

https://twiki.cern.ch/twiki/bin/view/Atlas/ImprovingSoftware

/usr/src/linux/Documentation/CodingStyle

http://atlas-sw.cern.ch/cgi-bin/viewcvs-atlas.cgi/offline/PhysicsAnalysis/JetTagging/JetTagCorrections/ImpactParameterTuning



http://atlas-sw.cern.ch/cgi-bin/viewcvs-atlas.cgi/users/sandvoss/DifferentialCrossSectionTTbar

http://atlas-sw.cern.ch/cgi-bin/viewcvs-atlas.cgi/users/sandvoss/DifferentialCrossSectionTTbar

162 BIBLIOGRAPHY

Fachbereich Mathematik und Naturwissenschaften Fachgruppe...

Documents

Transcript of Fachbereich Mathematik und Naturwissenschaften Fachgruppe...